Naqvi Et Al 2022 Decoding The Human Face Progress and Challenges in Understanding The Genetics of Craniofacial

Annual Review of Genomics and Human Genetics
Decoding the Human Face:

Progress and Challenges
in Understanding the
Annu. Rev. Genom. Hum. Genet. 2022.23:383-412. Downloaded from www.annualreviews.org
Genetics of Craniofacial
Access provided by 84.76.54.3 on 03/13/24. See copyright for approved use.
Morphology
Sahin Naqvi,1,2,∗ Hanne Hoskens,3,4,∗ Franziska Wilke,5
Seth M. Weinberg,6,7,8 John R. Shaffer,6,7 Susan Walsh,5
Mark D. Shriver,9 Joanna Wysocka,1,10,11,†
and Peter Claes3,4,12,13,†
1
Department of Chemical and Systems Biology, Stanford University School of Medicine,
Stanford, California, USA; email: naqvi@stanford.edu, wysocka@stanford.edu
2
Department of Genetics, Stanford University School of Medicine, Stanford,
California, USA
3
Center for Processing Speech and Images, Department of Electrical Engineering, KU Leuven,
Leuven, Belgium; email: hanne.hoskens@kuleuven.be, peter.claes@kuleuven.be
4
Medical Imaging Research Center, University Hospitals Leuven, Leuven, Belgium
5
Department of Biology, Indiana University–Purdue University Indianapolis, Indianapolis,
Indiana, USA; email: fwilke@iu.edu, walshsus@iupui.edu
6
Department of Human Genetics, University of Pittsburgh, Pittsburgh, Pennsylvania, USA;
Annu. Rev. Genom. Hum. Genet. 2022. 23:383–412 email: smwst46@pitt.edu, john.r.shaffer@pitt.edu
7
Center for Craniofacial and Dental Genetics, Department of Oral and Craniofacial Sciences,
First published as a Review in Advance on
University of Pittsburgh, Pittsburgh, Pennsylvania, USA
April 28, 2022
8
Department of Anthropology, University of Pittsburgh, Pittsburgh, Pennsylvania, USA
The Annual Review of Genomics and Human Genetics 9
Department of Anthropology, The Pennsylvania State University, University Park,
is online at genom.annualreviews.org
Pennsylvania, USA; email: mds17@psu.edu
https://doi.org/10.1146/annurev-genom-120121- 10
Department of Developmental Biology, Stanford University School of Medicine, Stanford,
102607 California, USA
11
Copyright © 2022 by Annual Reviews. This work is Howard Hughes Medical Institute, Stanford University School of Medicine, Stanford,
licensed under a Creative Commons Attribution 4.0 California, USA
International License, which permits unrestricted 12
Department of Human Genetics, KU Leuven, Leuven, Belgium
use, distribution, and reproduction in any medium, 13
Murdoch Children’s Research Institute, Melbourne, Victoria, Australia
provided the original author and source are credited.
See credit lines of images or other third-party
material in this article for license information
∗
These authors contributed equally to this work
†
These authors jointly supervised this work
383
Keywords
craniofacial, development, genome-wide association study, GWAS, syndromes, gene regulation,
phenotyping
Abstract
Variations in the form of the human face, which plays a role in our individual identities and so-
cietal interactions, have fascinated scientists and artists alike. Here, we review our current under-
standing of the genetics underlying variation in craniofacial morphology and disease-associated
dysmorphology, synthesizing decades of progress on Mendelian syndromes in addition to more
recent results from genome-wide association studies of human facial shape and disease risk. We
also discuss the various approaches used to phenotype and quantify facial shape, which are of par-
ticular importance due to the complex, multipartite nature of the craniofacial form. We close by
discussing how experimental studies have contributed and will further contribute to our under-
standing of human genetic variation and then proposing future directions and applications for the
field.
INTRODUCTION
The human face and craniofacial structures exhibit a high degree of variation both among in-
dividuals of our own species and in comparison with those of other great apes (106). Changes
in human craniofacial structures that occurred since the divergence from our closest living evolu-
tionary relatives, chimpanzees and bonobos, facilitated the emergence of the uniquely human facial
appearance and may have contributed to adaptations associated with bipedal locomotion, diet (79),
enlarged brain size (114), and speech articulation (91). Consistent with the rapid evolution of hu-
man facial traits, modern human skulls display characteristic features that are distinct from those
of extinct hominins such as Homo erectus, archaic humans, and Neanderthals (5), whereas other
facial attributes (e.g., aspects of nose shape) have been postulated to mediate climate adaptation
in human populations (174).
In addition to facilitating biological adaptations, the human face is a strong target of sexual
selection and plays key roles in communication and other social interactions (53, 138). Cranio-
facial malformations—which are among the most common congenital diseases—therefore have
not only huge medical implications but also devastating social consequences for affected patients
and their families. Understanding the factors driving human craniofacial variation in health and
disease is essential for comprehending what makes our faces both human and individual and for
developing therapeutic strategies, treatment plans, and reconstruction techniques for a myriad of
craniofacial disorders.
In this review, we discuss the contributions of genetics to understanding morphological cranio-
facial variation. We begin with a brief review of human craniofacial development, including some
insights from studies of other species, before discussing studies that explore the overall heritability
of human facial shape. Since the face is a complex morphological structure, we first discuss various
phenotyping approaches and then provide an overview of craniofacial dysmorphism in rare syn-
dromes. We next discuss genome-wide association studies (GWASs) of facial shape and associated
insights into the genetic architecture of common variation, as well as the interplay between com-
mon and rare variation and associated conceptual models. Finally, we discuss post-GWAS analyses
in the form of fine mapping and experimental functional studies before describing future possible
directions for the field.
384 Naqvi et al.

THE DEVELOPMENT OF THE FACE
The development of the face involves a series of highly coordinated embryonic events, most of
which occur between the third and eighth week of gestation (144). Tissues of the craniofacial com-
plex are derived primarily from cranial neural crest cells (CNCCs), a transient population of cells
that are specified from within the neural folds of the developing neural plate. Following specifica-
tion, CNCCs undergo an epithelial-to-mesenchymal transition and migrate ventrally, eventually
giving rise to most of the craniofacial skeleton and connective tissue through differentiation into
cartilage, bones, and tendons (29). The first major division of developing craniofacial tissues oc-
curs early in the fourth week of embryonic development, with the separation of CNCCs forming
the frontonasal prominence from those populating the four pharyngeal arches. The frontonasal
prominence later gives rise to the forehead and nasal bones (bridge and dorsum of the nose), while
the first branchial arch gives rise to the paired maxillary and mandibular prominences, which at this
point are still separated by the stomodeum (the precursor of the mouth). The remaining branchial
arches give rise to noncraniofacial structures. In the fifth week, nasal placodes, from which the lat-
eral nasal prominence and medial nasal prominence are derived, appear as thickenings on each
side of the frontonasal prominence, and the mandibular prominences merge at the midline, con-
tributing to the mandible, chin, and lower lip. Fusion of the medial nasal prominences is followed
by fusion with the maxillary prominences approximately seven weeks after conception, leading to
the formation of the central structures of the nose (columella) and central upper lip (philtrum).
The sides and alae (wings) of the nose form when the lateral nasal prominences and the maxillary
prominences begin to merge. These existing structures grow and mature during the remaining
weeks of pregnancy (145).
The complex morphogenesis of the face, from the early specification and migration of CNCCs
to the development of the facial prominences and derived structures, requires tightly controlled
spatial and temporal signaling networks and gene expression patterns. Many of the mechanisms
underlying early specification and morphological events have been uncovered by decades of study
in model systems in which the early developmental transitions observed in human embryos are
conserved, such as mouse and chick embryos. While a comprehensive review of such studies is
beyond the scope of this work, two key concepts from this field emerge: first, that both cell-
autonomous patterning of neural crest cells and embryo region-specific environmental cues by
signaling ligands are key for proper craniofacial development, and second, that different combi-
nations of the same signaling pathways pattern specific craniofacial regions in distinct ways. A ma-
jor mechanism by which this patterning is achieved is by the combinatorial expression, through
both cell-autonomous and signaling-responsive mechanisms, of sequence-specific transcription
factors. The gene regulatory networks formed by these transcription factors have been delineated
and reviewed in detail by Simões-Costa & Bronner (142). The major signaling pathways mediating
environmental cues to CNCCs and other craniofacial cell types include fibroblast growth factor
(FGF) (115), bone morphogenetic protein (BMP) (50), Sonic hedgehog (Shh) (41), wingless/Int-1
(128), and transforming growth factor beta (TGF-β) (40), all of which have been reviewed in de-
tail elsewhere. Despite complex interactions between CNCCs and other cell types, such as the
epithelium, in the formation of the craniofacial structures, cross-species transplantation experi-
ments in birds [reviewed by Schneider (133)] have demonstrated that species-specific features of
the craniofacial complex are driven largely by the neural crest, suggesting that CNCCs may in
turn be the major factor determining variation in craniofacial morphology and risk for disease in
humans, as discussed in subsequent sections.
Postnatally, the craniofacial complex undergoes continuous growth and remodeling, with the
peak of growth occurring at puberty (145). The precise timing tends to be different for males and
www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 385

females; facial maturity develops between 12 and 14 years of age in females and approximately
2 years later in males (80, 105). Such remodeling is generally driven by the balance between
osteogenesis, where osteoblasts form new bones and osteoclasts break down existing bone. These
patterns of controlled formation and destruction are driven at least partly by biomechanical
loading during normal function (43). The precise mechanisms by which these and other factors
contribute to variation in remodeling remain largely unknown. Recently, however, studies have
identified skeletal stem cells throughout the body (21, 22), and in the mouse lower jaw such
skeletal stem cells give rise to osteoblasts in response to surgical bone separation, likely going
through a neural crest–like intermediate state (126).
ESTIMATES OF FACIAL FORM HERITABILITY

Facial variation is determined largely by genetic variation in combination with diverse environ-
mental influences. The strong genetic component can be easily recognized and interpreted by per-
ceptual observation; for example, there is a strong resemblance of facial features across families,
and monozygotic twins are more alike than dizygotic twins. In addition, shared facial characteris-
tics within ancestries (81) and the sexes (84, 105) as well as the distinctive facial features associated
with particular genetic conditions (57) further demonstrate the extent to which genes influence
craniofacial form and appearance.
Craniofacial heritability studies have been performed to estimate the proportion of the phe-
notypic variation that can be explained by genetics, as compared with environmental effects. Tra-
ditionally, this is done by measuring the facial similarity between relatives, such as twins (39, 154,
162) or parents and offspring (60, 73). The widespread availability of whole-genome microar-
rays and recent advances in statistical methods allow for these traditional family-based designs
to be extended to large data sets of unrelated individuals where the degree of genetic relatedness
is determined from genome-wide single-nucleotide polymorphisms (SNPs) (28, 171). However,
these differences in study design, as well as differences in study populations and facial descrip-
tors employed, have produced widely varying estimates of heritability in the literature, making
direct comparisons across studies challenging. Research designs based on the similarity between
twins or siblings tend to overestimate heritability due to shared environmental sources of covari-
ance (common environment), whereas the lower bound is typically specified by the heritability
estimate obtained from common SNPs using approaches that assume only additive effects (44).
The acquisition of both genomic data and phenotypic measures for a sample of related individuals
poses an interesting avenue to dissect this bias. Furthermore, heritability is a dynamic rather than
a static descriptor, and estimates are expected to differ across populations and even within a single
population due to varying genetic and environmental influences (158).
Despite these differences, strong genetic influences have predominantly been found for cen-
tral midfacial parameters, with estimates exceeding 0.8 (i.e., >80% of variation is attributable to
genetics) (27, 60, 111, 154, 162). In particular, high heritability has consistently been reported for
aspects of nasal shape such as the position of the nasion, which is closely linked to the PAX3 locus
(1, 11, 19, 25, 61, 118, 164, 170). In addition to aspects of facial shape, facial size and allometry
have been found to be highly heritable (28, 39). By contrast, stronger environmental contributions
have been reported for the lower parts of the face, including the cheeks, mandible, and mouth,
which are known to be affected by nutrition (141), aging (119), and oral function (123). Other
nongenetic influences that can affect facial shape include prenatal exposure to teratogens [e.g.,
through maternal smoking (7) or alcohol intake (108, 150)], which may then lead to interactions
with genes involved in craniofacial development, and the effects of climate (174), geography (151),
and altitude (14) (e.g., through adaptation over evolutionary timescales).
386 Naqvi et al.

APPROACHES TO PHENOTYPING FACIAL FORM
Accurate and objective assessment of the human face, which is marked by inherent complexity
and multidimensionality, is essential to any study focusing on craniofacial variation. Methods de-
signed to measure aspects of facial form date back to at least the eighteenth century, with standard-
ized anthropometric methods emerging by the early twentieth century (63). The field has grown
exponentially since then through advances in acquisition techniques and analytical approaches.
The surface topography of the craniofacial complex is determined by two distinct but closely
related components: the underlying facial skeleton, or hard tissue, and the overlying soft tis-
sue facial structures (147). Compared with the skeletal component, the soft tissue component of
the face is affected more by parameters such as age and body mass index (86, 141). However, these
changes are unlikely to carry a strong craniofacial genetic component, and their effects are fairly
predictable, so they can be controlled for statistically by means of regression in studies focused on
the genetics of craniofacial variation.
Various imaging systems have been used to capture the facial surface and/or underlying skele-
ton in two or three dimensions (136). Three-dimensional imaging techniques are preferable be-
cause they successfully preserve the multidimensional nature of facial morphology. In facial genetic
research today, approaches based on stereophotogrammetry and laser scanning are commonly
used to generate high-quality images of the three-dimensional facial surface. Different acquisition
systems are commercially available, all of which are comparable in terms of accuracy, with error
rates well under 1 mm, as reviewed elsewhere (120). However, caution is needed when combining
data obtained from different imaging systems into a single analysis in order to avoid systematic
bias (165). An important drawback of three-dimensional imaging systems is that they are more
expensive than their two-dimensional counterparts and are not always portable, although this is
now beginning to change. Two-dimensional descriptions (1, 11, 19) and even categorical scoring of
facial features (1, 121) remain valuable, especially in the context of large-scale data collection, but
with the introduction of the TrueDepth camera in the iPhone X, three-dimensional facial imaging
is becoming more accessible to the larger public (131). In medical imaging, cephalometry can be
used to assess facial structures in two dimensions, while devices such as computed tomography
and cone-beam computed tomography can preserve the three-dimensional structure of the un-
derlying craniofacial skeleton in addition to the facial surface (136). However, radiation exposure
generally limits their availability to clinical settings. By contrast, magnetic resonance imaging is a
noninvasive technique that is directly informative about facial soft tissue structures, and it has been
adopted in large-scale projects such as the UK Biobank (148) and the Adolescent Brain Cognitive
Development (ABCD) study (17). The disadvantages of magnetic resonance imaging are its high
cost and long acquisition times.
Landmarks can be placed on two- and three-dimensional representations of a face from which
various quantitative measurements or features can be derived. Early approaches did so manually,
focusing on well-defined anatomical points, but extensive effort has been put into the development
of automatic methods to obtain fast, accurate, and reproducible results [4, 36, 90; for an in-depth
overview of different landmarking approaches, we refer readers to a paper by Böhringer & de
Jong (10)]. In addition, automatic methods allow for sparse descriptions to be expanded toward
spatially dense configurations of quasi-landmarks covering the entire facial surface (26, 66). Here,
the locations of the quasi-landmarks do not have any biological meaning per se, but homology is
implied by each quasi-landmark that occupies the same position on the face relative to all other
quasi-landmarks. Landmark configurations can subsequently be represented as linear distances,
angles, and ratios measured between the landmarks (traditional anthropometrics), or the land-
mark x-y-z coordinates themselves can be analyzed [geometric morphometrics (175)]. The latter

approach relies on superimposition methods to place the landmark configurations into a common
frame of reference (56).
Of primary interest in facial genetics is the variation within and among subjects. Dimension re-
duction techniques such as principal component analysis are frequently applied to decompose the
data into meaningful features while decreasing imaging noise. Indeed, the main modes of variation
in the data are effectively described by the first several principal components following principal
component analysis, and these modes have been analyzed separately or conjointly in genetic as-
sociation studies (25, 27, 93, 96, 118, 125, 164). Moreover, quantitative assessment of variation in
global facial shape has been complemented by the study of local features through the hierarchical
partitioning of facial shape into smaller segments using machine learning approaches (25). Alter-
natively, prior biological knowledge can be integrated to guide the facial features extracted [e.g., by
exploiting the facial resemblance within families (61)], or a lower-dimensional space can be built
using deep learning methods. Deep learning with neural networks is widely used in biomedical
image analysis (3), although it is fairly new to craniofacial genetic research, and most applications
still utilize two-dimensional data. For example, Face2Gene, a popular tool developed by FDNA
(Sunrise, Florida) to aid clinicians in the diagnosis of genetic disorders, applies deep learning algo-
rithms to two-dimensional photographs of patients to learn relevant facial features for syndrome
classification (54). Deep learning extensions for 3D surface or mesh data, referred to as geomet-
ric deep learning, are currently being developed (13, 101). However, the main challenge of deep
learning remains the large data burden. In addition, techniques are often tailored to solve specific
applications [e.g., facial classification (46)], and it is still unclear whether the learning outcomes
are also relevant biologically.
RARE VARIATION IN CRANIOFACIAL MORPHOLOGY

It has long been appreciated that craniofacial development can be severely disrupted. Since the
early twentieth century, clinicians have been defining such syndromes on the basis of their dis-
tinctive phenotypes, which often involve features beyond the face. Craniofacial syndromes are
individually rare but collectively common, with more than 500 clinical syndromes recognized
(143). Furthermore, while each syndrome has a distinct constellation of features, some are often
overlapping, including cleft lip and palate, craniosynostosis, and micrognathia.
Soon after the formalization of craniofacial syndrome diagnoses (37), and sometimes even
alongside them (31), it was observed that such syndromes are often inherited in a Mendelian
fashion, providing the first evidence for their genetic basis. At the same time, cases sometimes
arose sporadically with no family history. Both types of cases would later be useful for gene map-
ping efforts starting in the 1980s; for example, mapping the translocation breakpoints in sporadic
cases showed that GLI3 and SOX9 cause, respectively, Greig cephalopolysyndactyly syndrome
and campomelic dysplasia (which includes craniofacial malformations, referred to as Pierre Robin
sequence) (159, 160). By contrast, TCOF1 and FGFR2 were mapped as causative for Treacher
Collins and Crouzon syndromes, respectively, largely through linkage analysis in family pedigrees
(38, 127). Such gene mapping efforts were greatly accelerated by the development of Sanger se-
quencing and, later, massively parallel high-throughput approaches that allowed whole-exome and
whole-genome sequencing. As with many studies of rare variation, and in contrast to studies of
common variation described below, gene mapping was greatly simplified by the protein-coding
nature of most causative mutations.
To gain a systematic understanding of the processes impacted in craniofacial syndromes, we
searched the Online Mendelian Inheritance in Man (OMIM) database (58) for the term cranio-
facial and then manually reviewed each clinical synopsis to exclude erroneous results (i.e., when a
388 Naqvi et al.

lack of craniofacial syndromes was noted in the synopsis). This resulted in a list of 463 distinct clin-
ical synopses involving craniofacial defects, which mapped to 343 distinct protein-coding genes
(Figure 1; for further details, see Supplemental Table 1). Categorizing these genes into broad
functional classes indicates that, while a wide range of molecular and cellular processes are per-
turbed in craniofacial syndromes, certain functions are more represented than others. One notable
theme is the regulation of chromatin and transcription, both by sequence-specific transcription
factors (the largest single group) and by enzymes that modify and/or remodel chromatin more
broadly. Another prominent class is that of intracellular signaling ligands and their cognate recep-
tors, along with secondary regulators of the same pathways. The prominence of sequence-specific
transcription factors and signaling pathways can be understood in the context of the themes of
craniofacial development discussed above, where different combinations of intracellular signaling
ligands communicate environmental cues to cells of the developing face, in part through sequence-
specific transcription factors, which are also in turn responsible for more cell-autonomous pattern-
ing of the craniofacial structures. General chromatin modifiers, on the other hand, are expected to
have critical roles in the maintenance of transcriptional programs and cell identity in virtually all
cells of the body, raising the question of why they are so prominent in craniofacial syndromes. One
explanation is that some of these general regulators have additional, specialized functions that are
particularly important for cells such as CNCCs (15); another is that CNCCs and their derivatives,
along with other cell types that contribute to the face, are uniquely sensitized to these general per-
turbations of the chromatin and transcription machinery (52). Which of these two explanations is
more common remains to be seen. Nevertheless, the presence of such coherent functional classes
of genes and the large overlap with loci discovered by common variant association studies as well
as mouse models (discussed in subsequent sections) illustrate the overall success of gene mapping
rare craniofacial syndromes.
COMMON VARIATION IN CRANIOFACIAL SHAPE

Genetic Architecture
As discussed in the previous section, heritability studies of common facial variation have indicated
a large genetic contribution, but the underlying genetic architecture remained relatively unknown
until the advent of GWASs. While candidate gene studies have associated polymorphisms around
a preselected gene of interest with various measures of facial shape (24, 68, 130), such studies do
not provide insights into the overall distribution of variants affecting facial shape. GWASs test
a panel of variants (typically SNPs) across the genome for association with a phenotype (here a
facial shape measure); the first facial shape GWAS appeared in 2012. To date, there have been
25 GWASs of facial shape; they are summarized in Table 1, with additional details on loci from
each study provided in Supplemental Table 2. As with most GWASs, the sample sizes started
out small (several hundred individuals) and have since become much larger (tens of thousands of
individuals), although these GWASs are not as large as many others for several reasons, includ-
ing the limited availability of both genomic data and accurate facial scans from large numbers of
individuals and the privacy issues inherent in obtaining such data on a large scale.
The number of loci discovered to date suggests that facial shape as a trait is highly polygenic
in nature, but a more precise understanding of its genetic architecture requires quantifying
the effect size of associations. This is something of a challenge for facial shape, as different
studies have used different phenotyping approaches, some univariate and some multivariate. We
aggregated all independent SNPs previously associated with facial shape from published GWASs
(Supplemental Table 2) and used partial least-squares regression to estimate the fraction of
variance explained by all such loci in a sample of 4,680 European-ancestry individuals (164). We

390
ANKH BCOR ACAN CWC27 ANKRD11 ALX1 PAX3 BMP1 AKT1 ALDH18A1 COLEC10
ATP6V0A2 LZTR1 ANTXR1 PARN ARID1B ALX3 PAX6 BMP4 ARHGAP31 B3GALT6 COLEC11
CCBE1 PQBP1 ASXL1 ALX4 PBX1 BMPER BRAF B3GAT3 CPLANE1
Naqvi et al.
IPO8 MBD5
CDH1 CSF1R CHST3 DYNC1I2
KCNH1 RAI1 RBMX ATRX ARX PRRX1 CHUK
CDH2 CTNNB1 CHST14 EPB41L1
KCNJ2 SNIP1 SF3B4 BICRA BCL11B RAX EDN1 FGD1
CDH11 BPTF DLX3 RUNX2 CYP26B1 FLNA
KCNMA1 TCOF1 SMG9 EDNRA GNAS
COL1A1 DHCR7 FLNB
regulator
PAM16 ZMIZ1 SNRPN CHD7 DLX5 SATB2 EFNB1 HRAS
COL1A2 DPYSL5 IFT43
processing
EED ERF SIX3 ERBB3
RNA binding,
SLC10A7 ZMYM2 COL11A1 SPEN KRAS EBP LMNA
THOC6 EP300 FOXC1 SOX2 EVC2 MAPK1
SLC25A24 ZMYND11
Other transcriptional
DCHS1 FGFR1 ESCO2 MAPRE2
SLC35D1 FAM20C WBP11 EZH2 FOXL2 SOX9 FGFR2 MAP2K1 FTO MID1
TMCO1 FBN1 HDAC8 GATA6 TBX1 FGFR3 MAP2K2 GALNT2 NHS
FRAS1 CDC6 KAT5 GLI2 TBX2 FZD2 MTOR HCCS PCLO
ASPH FREM1 CDC45 KAT6B GLI3 TBX15 GDF5 PTPN11 HHAT RHOA
FBXL4 DPH1 GJA1 DDX11 KDM1A GSC TBX22 GH1 RAF1 HS2ST1 RTTN
GPC4 IL11RA HSD17B4 SPECC1L
Intracellular structure
HSPA9 GNPTAB FANCA KDM6A IRF6 TCF12 RIC1

LRPPRC HSPG2 JAG1 HYAL1 TAPT1
GUSB GMNN KMT2D IRX5 TFAP2A
General signal transduction
RIPK4
replication
LRP4
DNA repair,
LTBP3 INPPL1 TNNI2
ITCH MGP ORC1 MORC2 MAF TFAP2B NOTCH2 RIT1
Transporter, ion channel Mitochondria

MAB21L1 TOGARAM1
ADAMTS2 OTUD5 MMP2 NIPBL MEF2C TGIF1 NOTCH3 SHOC2
Metabolic enzymes
MAB21L2 TPM3
ADAMTS10 SUMO1 MMP14 ACTB NSD1 MEIS2 TP63 PTCH1 SOS2 NAA10 TRIO
modification
Extracellular matrix, cell adhesion
TRIM37 MYMK CPLANE1 PCGF2 MEOX1 TRPS1 PTCH2 ZPR1
ECE1 PTCH3 NAA15 WDPCP
KLHL7 UBE3B PKDCC DYNC1I2 PHF8 MITF TWIST1 PIGL
Chromatin modification, remodeling

PTH1R
Sequence-specific transcription factor
Other posttranslational
LONP1 PLOD3 EPB41L1 PHF21A MSX1 TWIST2 PIK3CA
PTPRF
RELN FLNA
Signaling ligand/receptor
MASP1 SMARCA4 MSX2 VSX1 ROR2 PLCB4 EFTUD2

SCUBE3 FLNB SMARCC2 NFIA VSX2 RSPO2 PRUNE1
PSMD12 EIF4A3
SH3PXD2B
Proteolysis
PEX1 IFT43 SMARCE1 NKX2-5 ZBTB20 SEMA3E ARCN1 PSAT1 IARS2
TASP1
PEX2 MAPRE2 SRCAP NKX2-6 ZEB2 SHH COG4 PTDSS1 RPL5
UBR1 SKI PYCR1
ZMPSTE24 PEX3 CENPF MID1 STAG2 OTX2 ZNF462 IFT172 RPL11
SMO SC5D
PEX5 CENPT NHS SOST PACS1 RPL15
TXNL4A
PEX6 CEP57 PCLO INTS1
Cytoskeleton
STRA6 SEC23A RPL35A
CCDC32 PEX10 CEP120 RTTN MED12 CDC6 SUFU SEC24D RPS7
MN1 PEX13 CKAP2L TAPT1 MED13L CDC45 TGFB3
Translation
SEC31A CDCA7 RPS10

TGFBR1
Peroxisome
RSPRY1 PEX14 KIAA0753 TNNI2 POLR1A DDX11
chaperones
TBCE CDK13 RPS17

TMEM94 PEX16 SMC1A TOGARAM1 POLR1B FANCA TGFBR2
VPS13B CDKN1C RPS19
Chromosome
WNT3
machinery
WDR37 PEX19 SMC3 TPM3 POLR1C GMNN
Unknown
VPS35L NDN RPS24
replication
WNT5A
DNA repair,
Cell cycle
Intracellular trafficking,
regulation
ZSWIM6 PEX26 SMCHD1 WDPCP POLR3A ORC1 WNT10B WASHC5 NEK1 RPS26
organization, segregation
General transcription
Figure 1
Genes and pathways mutated in rare craniofacial disorders with Mendelian inheritance. The genes were identified by curating hits for the search term craniofacial from
the OMIM database, which were then organized into broad functional categories based partly on their PANTHER protein classes. Associations with GWASs were
defined by aggregating the candidate genes and loci from studies listed in Table 1. Genes that cause craniofacial phenotypes when mutated in mice (red) were found by
querying the Mouse Phenome Database. Blue shading indicates genes that GWASs have implicated in facial shape or nsCL/P. Abbreviations: GWAS, genome-wide
association study; nsCL/P, nonsyndromic cleft lip with or without cleft palate; OMIM, Online Mendelian Inheritance in Man; PANTHER, Protein Annotation Through
Evolutionary Relationship.
Table 1 GWASs of facial shape in healthy individuals

Study Year Study design Ancestry Phenotypes Key findings
Paternoster et al. (118) 2012 Discovery (N = 2,185), European Linear distances (N = 54) and principal Identified genome-wide significant
replication (N = 1,622) components (N = 14) derived from associations at 4 loci; replicated 1
three-dimensional facial images signal; candidate gene PAX3
Liu et al. (96) 2012 Discovery (N = 5,388), European Linear distances (N = 36), principal Identified genome-wide significant
replication (N = 4,435) components (N = 11), and centroid associations at 5 loci; candidate
size derived from three-dimensional genes PRDM16, PAX3, TP63,
head magnetic resonance imaging and C5orf50, and COL17A1
two-dimensional facial images
Jacobs et al. (70) 2014 Meta-analysis (N = 6,631) European Graded categorical scoring of sagging Identified genome-wide significant
eyelids (N = 1) associations at 1 locus; candidate
gene TGIF1
Pickrell et al. (121) 2016 Discovery (N = 70,000) European Diverse traits and diseases (N = 42), Identified genome-wide significant
including the (graded) categorical associations at 13 and 57 loci
facial features (N = 2) nose size and affecting nose size and chin dimples,
chin dimples respectively
Adhikari et al. (1) 2016 Discovery (N = 6,275), Latin American Linear distances and angles (N = 10) Identified and replicated genome-wide
replication (N = 501) and graded categorical features significant associations at 6 loci;
(N = 14) derived from two- and candidate genes EDAR, PAX3,
three-dimensional facial images DCHS2, SUPT3H/RUNX2, GLI3,
www.annualreviews.org
and PAX1
•
Shaffer et al. (137) 2016 Meta-analysis (N = 3,118) European Linear distances (N = 20) derived from Identified genome-wide significant
three-dimensional facial images associations at 7 loci; candidate
genes MAFB, PAX9, MIPOL1,
ALX3, HDAC8, and PAX1
Cole et al. (27) 2016 Discovery (N = 3,505), African Linear distances (N = 25), principal Identified and replicated genome-wide
replication (N = 2,390) components (N = 6), and facial size significant associations at 2 loci;
measures (N = 3) derived from candidate genes SCHIP1 and
three-dimensional facial images PDE8A
Lee et al. (88) 2017 Meta-analysis (N = 2,817) European Factors of linear distances (N = 23) Identified genome-wide significant
derived from three-dimensional facial associations at 3 loci; candidate
images genes PARK2 and FREM1
Crouch et al. (30) 2018 Discovery (N = 1,832), European Principal component extremes (N = 40) Identified genome-wide significant
replication (N = 1,567) derived from three-dimensional facial associations at 27 loci; replicated 3
images signals; candidate genes PCDH15,
MBTPS1, and TMEM163
Genetics of Craniofacial Morphology and Disease

Claes et al. (25) 2018 Discovery (N = 2,329), European Combined principal components Identified genome-wide significant
replication (N = 1,719) representing global-to-local facial associations at 38 loci; replicated 15
391
segments (N = 63) derived from signals; candidate genes TBX15,
three-dimensional facial images ASPM, PKDCC, HOXD cluster,
PAX3, RAB7A/ACAD9, EPHB3/
DVL3, DCHS2, SUPT3H, RPS12/
EYA4, DLX6/DYNC1L1, BC039327,
SOX9, and KCTD15
(Continued)
Table 1 (Continued)
Endo et al. (42) 2018 Meta-analysis (N = 11,311) East Asian Diverse skin-related traits (N = 7), Identified genome-wide significant
including the categorical scoring of associations at 1 locus affecting
double-edged eyelids double-edged eyelids; candidate
genes EMX2, and EMX2OS
Cha et al. (19) 2018 East Asian Linear distances, angles, areas, and Identified genome-wide significant
392
Discovery (N = 5,643),
replication (N = 1,926) curvature (N = 85) derived from associations at 5 loci; candidate
two-dimensional facial images genes OSR1-WDR35,
HOXD1-MTX2, WDR27, SOX9,
and DHX35
Howe et al. (62) 2018 Meta-analysis (N = 6,136) European Linear distance (N = 1) derived from Identified genome-wide significant
Naqvi et al.
three-dimensional facial images associations at 2 loci affecting
representing philtrum width philtrum width; candidate genes
HOXA1 and ENSG00000232633
Qiao et al. (125) 2018 Discovery (N = 1,709), European, East Linear distances (N = 10), principal Identified genome-wide significant
replication (N = 1,675) Asian, admixed components (N = 6), and partial associations at 6 loci; replicated 4
European– least-squares components (N = 6) signals; candidate genes COL23A1,
Asian derived from three-dimensional facial PCDH7, UBASH3B, and BMP2
images
Wu et al. (169) 2019 Discovery (N = 50) East Asian Linear distances (N = 48) derived from Identified genome-wide significant
three-dimensional computed associations at 4 loci; candidate
tomography scans of the skull genes RGPD3, IGSF3, SLC28A3,
and USP40
Xiong et al. (170) 2019 Discovery (N = 10,115), European, Latin Linear distances (N = 78) derived from Identified genome-wide significant
replication (N = 7,917) American, three-dimensional facial images associations at 24 loci; replicated 10
admixed signals; candidate genes CASZ1,
European– INTU, KIF6, TBX3, C14orf64,
Asian RPGRIP1L, PAX3, SFRP2, ROR2,
and PAX1
White et al. (164) 2021 Meta-analysis (N = 8,246) European Combined principal components Identified genome-wide significant
representing global-to-local facial associations at 203 loci and
segments (N = 63) derived from study-wide significant associations
three-dimensional facial images at 120 loci; found 117 new and 86
previously identified signals
Bonfante et al. (11) 2021 Discovery (N = 6,169) Latin American Linear distances, ratios, and angles Identified genome-wide significant
(N = 59) derived from associations at 32 loci; replicated 23
two-dimensional facial images signals and found 9 new signals
Indencleef et al. (67) 2021 Meta-analysis (N = 8,246) European Scores representing nsCL/P Identified genome-wide significant
endophenotype (N = 59) defined associations at 29 loci and
across global-to-local facial segments study-wide significant associations
(N = 63) derived from at 9 loci; 22 signals were previously
three-dimensional facial images identified in GWASs on normal-
range facial variation, and 18 signals
were near genes with strong
evidence in orofacial clefting
(Continued)
Table 1 (Continued)
Huang et al. (65) 2021 Discovery (N = 2,659); East Asian Combined principal components Identified genome-wide significant
additionally tested representing different facial segments associations at 7 loci and study-wide
56 SNPs previously (N = 10) derived from significant associations at 4 loci;
reported in facial GWAS three-dimensional facial images replicated 24 SNPs from previously
literature reported genetic loci near DCHS2,
SUPT3H, HOXD1, SOX9, PAX3,
and EDAR; candidate genes
DENND1B, PISRT1, DCHS2/
SFRP2, and VPS13B
Liu et al. (94) 2021 Discovery (N = 2,329); European Combined principal components Identified significant associations at
tested low-frequency representing global-to-local facial 7 genes: HFE, NECTIN1, CARS2,
coding variants segments (N = 31) derived from LTB4R, TELO2, AR, and FTSJ1
(MAF < 1%) in 8,091 three-dimensional facial images
genes
www.annualreviews.org
Hoskens et al. (61) 2021 European Scores representing facial traits shared Identified genome-wide significant
•
Meta-analysis (N = 8,246)
by siblings (N = 1,048) defined across associations at 218 loci and study-
global-to-local facial segments wide significant associations at 38
(N = 63) derived from loci; found 109 new and 109
three-dimensional facial images previously identified signals
Liu et al. (93) 2021 Discovery (N = 2,595) African Combined principal components Identified genome-wide significant
three-dimensional facial images at 6 loci; 10 signals were shared with
Europeans
Wang et al. (161) 2021 Meta-analysis (N = 9,674) East Asian Combined principal components Identified genome-wide significant
three-dimensional facial images at 151 loci; found 130 new and 114
previously identified signals; 89
signals were shared with Europeans
Genetics of Craniofacial Morphology and Disease

Knol et al. (85) 2021 Meta-analysis European Interorbital distance derived from Identified genome-wide significant
(N = 34,130), replication magnetic resonance imaging associations at 56 loci; 46 signals
(N = 10,115) were new for facial morphology
393
Abbreviations: GWAS, genome-wide association study; MAF, minor allele frequency; nsCL/P, nonsyndromic cleft lip with or without cleft palate; SNP, single-nucleotide polymorphism.
found that 501 independent SNPs (r2 < 0.1), encompassing 303 loci, explained 13.7% of variance
of the full face, represented by the first 36 principal components as determined by parallel analysis.
In the same cohort, age, sex, and body mass index individually explained 7.0%, 12.2%, and 18.9%
of the variance, respectively. These results are consistent with a polygenic architecture typical of
complex traits, where many variants contribute individually small effects. As a comparison, for
human height (one of the most polygenic complex traits), 3,290 independent, genome-wide sig-
nificant variants explain 24.6% of variance (172). Despite having a polygenic architecture overall,
a small number of common variants can have quite large effects on facial shape; a recent study
specifically comparing the upper and lower tails of the distribution of facial shape (measured by
specific principal components describing shape variation) found three variants with frequencies
of approximately 10% and unusually large effects (odds ratios >4 for being in either tail) (30).
The two most common craniofacial birth defects, cleft lip with or without cleft palate (CL/P)
and craniosynostosis, are an interesting point of contrast to the architecture of normal-range

facial shape variation. As discussed above, both of these defects are found in a number of Mendelian
craniofacial syndromes, but approximately 70% (74) and 50% (140) of CL/P and craniosynostosis
cases, respectively, do not appear as part of a syndrome. Nonsyndromic CL/P (nsCL/P) occurs
in 1 of every 920 live births in the United States (117), while the prevalence of isolated cran-
iosynostosis has been estimated at between 1 in 2,100 and 1 in 2,500 live births (12, 87). There
have been numerous GWASs of nsCL/P (6, 8, 35, 51, 64, 89, 98–100, 102, 103, 107, 149); these
studies have been reviewed in detail elsewhere (152), but we also provide a combined list of vari-
ants and loci discovered by these studies in Supplemental Table 3. Together, these studies have
demonstrated a substantial contribution of common variants to nsCL/P risk, with one study find-
ing that approximately 30% of the variance in nsCL/P is attributable to all common SNPs (99).
Notably, the same study found that the top 24 loci explain 25% of nsCL/P risk (99) in European-
ancestry individuals, and another found that 26 loci explain approximately 11% of the variance
in a Chinese cohort (173), indicating substantially larger effect sizes for nsCL/P than for facial
shape variation. While nonsyndromic craniosynostosis common-variant GWASs have had less
statistical power for variant discovery due to their smaller sample sizes, variants at the BMP2 lo-
cus have been independently associated with two main subtypes, sagittal (76) and metopic (75)
craniosynostosis. Here, too, the effect sizes of lead variants are unusually large relative to both
facial shape variation and most common diseases (odds ratios >4 for sagittal and >2 for metopic
craniosynostosis).
While the vast majority of GWASs to date have sought to discover variants with additive ef-
fects on the phenotype of interest, genetic interactions—either with other variants (gene × gene
interactions, or epistasis) or with environmental variables—can be an additional source of vari-
ation. As a whole, GWASs of diverse phenotypes have found very few examples of large-effect
epistatic interactions (129), and facial shape GWASs are similar in that regard. Several general
explanations have been provided for this lack of epistasis, including a lack of statistical power to
detect epistatic effects (as they are far smaller than the main additive effects) and the testing of
tagging SNPs rather than the causal variant itself. Nevertheless, a few exceptions do exist. White
et al. (164) found four pairs of loci, each with its own additive associations with distinct multi-
variate facial phenotypes, which show significant epistasis. Furthermore, Liu et al. (95) described
an interaction between the PRICKLE1 and FOCAD loci for cranial base width underlying a vari-
ance quantitative trait locus (QTL) at PRICKLE1. With respect to craniofacial disease, three dis-
tinct modifiers of nsCL/P subtypes have been discovered in the last few years: the FAT4 locus
for cleft laterality (34), the PAX1 locus for bilaterality versus unilaterality (33), and a third locus
for cleft palate (16).
394 Naqvi et al.

Another important aspect of genetic architecture involves a description of genetic effects as
a function of genetic ancestry. As with most GWASs to date, most facial shape GWASs have
been performed in individuals of primarily European ancestry, making systematic comparisons
across populations challenging. However, some notable studies in non-European individuals have
included Latin Americans with admixed ancestry, Han Chinese individuals, and East African in-
dividuals (Table 1). While a direct comparison of phenotypic effects is somewhat challenging
since the phenotype itself (i.e., the average facial shape) can be substantially divergent between
populations, these studies have revealed that most of the variants affecting facial shape are shared
between populations but a subset are likely population specific. This was clearest for East African
individuals, which is unsurprising due to the higher divergence in allele frequency and linkage
disequilibrium between African and non-African populations (93). Furthermore, a recent GWAS
of nsCL/P risk from multiple broad ancestry groups found substantial dependence on ancestry in
terms of SNP effect size and significance (109).

Using Genome-Wide Association Studies to Understand Craniofacial Biology

In addition to describing genetic architecture (i.e., the number and effect sizes of variants asso-
ciated with facial shape), GWAS approaches can also be used to unravel the biology underlying
human facial variation in two related ways: first, by prioritizing specific cell types and cis-regulatory
element classes important for variation, and second, by nominating candidate genes and pathways
involved in the formation of the face.
As the vast majority of GWAS loci lie in noncoding regions of the genome, it follows that
they lie in (or tag causal variants lying in) cis-regulatory elements controlling expression of coding
genes. The cell types or tissues that are most enriched for GWAS-tagged regulatory elements
would then be predicted to have important roles in trait variation. To date, maps of regulatory
elements have been produced from both in vitro stem cell–derived CNCCs and their further
derivatives (i.e., chondrocytes) (97, 122), as well as primary embryonic craniofacial tissue from
Carnegie stages 12–20, during which time the craniofacial structures are still forming (167). The
most straightforward and common approach uses a predefined set of facial GWAS loci (typically
genome-wide significant SNPs) and either computes the amount of signal of a certain chromatin
feature [such as acetylation of lysine 27 on histone H3 (H3K27ac)] within a set genomic distance of
these variants or calculates the frequency with which the variants lie in regulatory elements defined
by combinations of chromatin features. These enrichments or frequencies for the set of GWAS
loci can then be compared with enrichments from a background set of loci not associated with
facial shape to estimate the significance of enrichment. Such an approach has been applied to both
normal-range facial GWAS loci (25) and loci associated with nsCL/P (99, 163) and has consistently
found significant enrichment or overlap with active regulatory elements (mostly enhancers rather
than promoters) in both in vitro–derived CNCCs and embryonic craniofacial tissue, as compared
with other embryonic and adult human cell types and tissues that have been examined.
A newer, complementary type of approach incorporates genome-wide summary statistics along
with patterns of linkage disequilibrium to compute an overall heritability enrichment of a GWAS
for a specific cell type annotation (45). Although such an approach can have advantages over the
simpler, cutoff-based approach, tools using it have generally been developed to work with highly
powered GWAS and univariate traits, making it challenging to apply to facial shape GWASs, which
are often multivariate in nature and use relatively small sample sizes. However, a recent study of
the shared genetics of face and brain shape variation showed that the most widespread of these
tools, linkage disequilibrium score regression, can be extended to the multivariate space with a
few modifications. In addition to revealing a significant enrichment of face shape heritability in

the in vitro–derived CNCCs, their chondrocyte derivatives, and embryonic craniofacial tissues,
this approach showed that these cell types were not enriched for brain shape heritability (112).
Overall, such studies suggest that much of the variation in facial shape arises from variants acting
in a range of cell types during early embryonic development, although epigenomic maps of many
of the human cell types and time points that are likely important to craniofacial development
remain incomplete.
With respect to nominating candidate genes, one issue, which we discuss more below, is that
both identification of the causal variant and assignment of the gene affected by the variant (since
the vast majority of GWAS loci lie in noncoding regions) are far more challenging than they are
for most craniofacial syndromes, which involve protein-coding variants. A common approach is
to assign a GWAS-associated SNP to either the nearest gene or a gene within a fixed genomic
distance that has known roles in craniofacial biology. While clearly heuristics, such approaches
have revealed that many facial GWAS-associated loci contain genes implicated in rare cranio-
facial syndromes discussed above (Figure 2). For example, the first GWAS locus associated with
common facial variation was PAX3, rare variants in which are known to cause Waardenburg
syndrome. Additional candidate genes highlighted by GWASs include SOX9, GLI3, TWIST1,
ALX1/4, MSX1, TFAP2B, DLX5/6, FGFR2, and BMP2/4. Indeed, of the craniofacial syndrome
genes also highlighted by GWASs, there is a striking enrichment for sequence-specific transcrip-
tion factors (Figure 1). There are several possible explanations for such an enrichment, ranging
from the broad regulatory but cell type–specific roles of transcription factors to their inherent
dosage sensitivity. Such hypotheses would be best addressed with genome-wide mapping of tran-
scription factor targets as well as experimental measurements of their dosage sensitivities in the
relevant cell types, which are yet to be done systematically.
INTERPLAY BETWEEN COMMON AND RARE VARIATION

Examples from Cleft Lip and Palate, Craniosynostosis, and Rare-Variant
Studies of Healthy Individuals
Genetic studies suggest that both common and rare variation in craniofacial shape lie on the same
continuum and share the same biological pathways, but to different extents: Common genetic
variation results in subtle and cell type–specific changes in gene expression, largely by modulat-
ing activities of cis-regulatory elements such as enhancers, whereas rare variation results in larger
changes in expression or activity of the same genes by impacting protein-coding sequence directly.
One model to synthesize these observations is a liability threshold model, where a certain outcome
(i.e., disease) occurs only after an individual passes a threshold due to the sum of multiple indepen-
dent factors. The liability threshold model for disease has been appreciated for decades, but direct
evidence that both common and rare genetic variation operate through the same pathways has
emerged only in more recent years. For example, common genetic variants can modulate the risk
of severe rare neurodevelopmental disorders, even though such disorders are largely monogenic
and caused by rare (typically protein-coding) mutations (116). At the molecular level, common
cis-regulatory variants have been observed to modify disease penetrance by either downregulating
expression of rare pathogenic coding variants in healthy individuals or upregulating their expres-
sion in diseased individuals (18).
While most of the craniofacial disease burden is likely due to rare genetic variation, an inter-
esting exception is CL/P, which, as discussed in the section titled Rare Variation in Craniofacial
Morphology, is a phenotypic feature of many different craniofacial syndromes. However, a major-
ity (approximately 70%) of CL/P cases are nonsyndromic, meaning that they occur in the absence
of other clinical features. While nsCL/P may have a somewhat different genetic architecture from
396 Naqvi et al.

Normal
displacement
Minor allele
Major allele
PAX3 TWIST1 ALX1 SOX9

rs1370926 rs212672 rs11609649 rs9908442
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 X
Figure 2
Genomic locations and regional effects of genome-wide significant loci for facial shape. In the ideogram, the
genomic locations of lead SNPs identified through GWASs of facial shape in healthy individuals (listed in
Supplemental Table 2) are shown in blue, and genes implicated in rare craniofacial disorders with
Mendelian inheritance (listed in Figure 1) are shown in orange. Associated phenotypic effects of the PAX3
(lead SNP rs1370926), TWIST1 (lead SNP rs212672), ALX1 (lead SNP rs11609649), and SOX9 (lead SNP
rs9908442) loci are illustrated by the facial morphs, exaggerated in the direction of the minor and major
allele SNP variant based on the results of White et al. (164). Heat maps represent the normal displacement
(displacement in the direction locally normal to the facial surface) in each quasi-landmark going from the
minor to the major allele SNP variant. Blue indicates inward depression; red indicates outward protrusion.
Abbreviations: GWAS, genome-wide association study; SNP, single-nucleotide polymorphism.
normal-range facial shape variation with respect to effect sizes, many loci are associated with both
phenotypes. Studies explicitly testing the effects of nsCL/P risk loci on facial shape have found
multiple loci with highly significant associations on both (9, 68). Indeed, one study found a sig-
nificant association between a polygenic risk score for nsCL/P and the width of the philtrum,
disruption of which leads to nsCL/P (62). Another recent GWAS of healthy, unaffected individu-
als, which used endophenotypes (intermediate phenotypes that can be used as markers for disease
risk) derived from directly comparing facial scans of unaffected relatives of patients with nsCL/P
with those of controls, found both overlap with the results of previous facial shape and nsCL/P

GWASs and associations between polygenic nsCL/P risk scores and several endophenotypes (67).
Thus, the case of nsCL/P suggests a liability threshold linking both normal-range and syndromic
variation in facial shape.
If both common and rare variants operate along the same axis in craniofacial shape and dis-
ease, one would also expect epistasis between common and rare variation with respect to dis-
ease presentation, since common variants could move certain individuals away from or toward
the threshold beyond which disease manifests. Exactly this type of interaction has been observed
for nonsyndromic craniosynostosis, for which exome sequencing and analysis of de novo muta-
tions in parent–child trios found a strong yet incompletely penetrant association with mutations
in SMAD6, a negative regulator of BMP-dependent osteoblast differentiation (153). Remarkably,
much of this incomplete penetrance was explained by a strong epistatic interaction with a common
variant at the BMP2 locus, which was discovered to have an independent effect in prior GWASs.
Furthermore, this interaction occurred in the same pathway: BMP ligands bind and induce their
cognate receptors to phosphorylate SMAD and activate transcription of osteogenic genes, a pro-
cess directly inhibited by SMAD6.

Another test of a common axis between rare and common variants is to assess the impact of
rare variants on facial morphology, but in individuals free of craniofacial disease. Rare variants
are naturally expected to have larger effect sizes when detectable, and they can therefore bridge
some of the gap between results from GWASs focusing on common variants and craniofacial
disorders (syndromic or nonsyndromic). Few studies have assessed the impact of rare variants
on facial shape due to the larger sample sizes required, but a recent study that focused on vari-
ants with <1% frequency and performed gene-based tests found seven genes to be enriched for
rare variants affecting facial shape (94). One such gene, NECTIN1, has known roles in cranio-
facial development and is associated with a syndrome (cleft lip/palate–ectodermal dysplasia syn-
drome), but, interestingly, the authors found no evidence of common variant associations within
or near these seven genes. While this may suggest that rare variation can impact pathways discrete
from those impacted by common variation, these low-frequency variants could still have epistatic
interactions with common variants that could not be detected at the study’s sample size (2,329
individuals).
A Multivariate Model Linking Craniofacial Shape Variation and Disease

Studies to date have provided evidence of a liability threshold model for nsCL/P risk and spe-
cific univariate measures of facial shape, as well as clear epistatic interactions between rare and
common variants for craniosynostosis (Figure 3a). But how do these findings generalize to the
entirety of common facial shape variation, which GWASs have shown to be under multivariate
genetic control, and craniofacial disorders, which range from the fairly prevalent and polygenic
(such as nsCL/P and craniosynostosis) to Mendelian syndromes caused by one or a handful of
genes (such as Treacher Collins and Crouzon syndromes) (Figure 3b), each of which has variable
yet distinct facial features? We propose a description of craniofacial variation involving a multi-
variate shape space, a two-dimensional schematic of which is portrayed in Figure 3c. In this model,
an individual lies at a point in high-dimensional space, the axes of which can be thought of as dif-
ferent aspects of shape variation. The distribution of individuals in this space can be thought of as
a multivariate Gaussian distribution, and both genetic (common or rare) and environmental fac-
tors can move individuals around in this high-dimensional space. Different disease states occupy
distinct but potentially overlapping zones in the space. For example, Mendelian craniofacial syn-
dromes would occupy largely distinct zones due to their distinct facial characteristics, but nsCL/P
and craniosynostosis would occupy a zone that overlaps multiple syndromes (i.e., Pierre Robin
398 Naqvi et al.

a Cleft lip and palate b
0.4 Cleft lip
and palate
Density of individuals
Disease prevalence
0.3
Craniosynostosis
Stickler
0.2 Saethre– Pierre
Chotzen Robin
Treacher
0.1 Collins
Crouzon
0.0
Cleft lip and palate liability Low High
Polygenicity
Genetics,
environment Cleft lip and palate liability
Philtrum width
Treacher
Collins Stickler
Pierre
2
Cleft lip and palate Robin

n
nsio
Genetics,
environment
ime
Saethre–
pe d
Chozen
Sha
Craniosynostosis
Crouzon
Shape dimension 1
Figure 3
A multivariate model linking craniofacial shape and disease. (a) In standard liability threshold models, disease risk, which can be
modulated by genetics or the environment, manifests itself as a univariate trait, or endophenotype (e.g., philtrum width for
nonsyndromic cleft lip and palate). (b) Craniofacial disorders have a range of genetic architectures, from highly polygenic (right) to
monogenic with largely Mendelian inheritance (left). (c) In the proposed multivariate shape space model, variation along multiple axes
of shape (only two of which are shown for visualization purposes) leads to distinct yet overlapping zones of disease in the
multidimensional shape space. The craniofacial syndromes are arranged approximately according to phenotypic similarity. Facial
morphs for syndromes were created by first creating a univariate shape score distinguishing individuals with the syndrome from healthy
controls and then moving four standard deviations along the univariate syndromic axis.
sequence, Treacher Collins syndrome, and Stickler syndrome for nsCL/P and Saethre–Chotzen
and Crouzon syndromes for craniosynostosis). Such a model can also accommodate the variable
expressivity observed in many Mendelian syndromes, since additional factors, such as an individ-
ual’s polygenic background or developmental environment, could shift them away from the center
of the shape space zone corresponding to that syndrome.

FUNCTIONAL STUDIES
There have been relatively few experimental functional studies directly based on findings from
human genetic studies of craniofacial morphologies, in part because it can be challenging from
both an ethical and practical perspective to re-create much of human craniofacial development in
a dish. However, developments in human stem cell differentiation approaches, greater access to
primary tissues from early developmental time points, and the use of mouse models have begun
to provide some insights in this relatively nascent field. Here, we highlight some of the major
objectives of such studies and a few examples.
Experimental Fine Mapping and Elucidating Target Genes

Due to linkage disequilibrium between variants throughout the genome, population-based associ-
ation studies such as GWASs inherently highlight regions of the genome and sets of linked SNPs
rather than exact causal variants underlying a trait. Numerous approaches, collectively termed fine
mapping, have been used to further narrow down candidate causal variants from GWASs. These
include cross-ancestry GWASs, where distinct patterns of linkage disequilibrium between popu-
lations can narrow down candidate variants (assuming the same causal variant underlies variation
in both populations); targeted fine-mapping GWASs, with denser panels of variants that allow
for imputation of smaller linkage blocks; and purely computational approaches. These statistical
approaches have been reviewed in detail by Schaid et al. (132). Here, we focus on fine-mapping
approaches based on experimental data.
Experimental fine-mapping approaches vary in both the strength of evidence provided and
scalability. Perhaps the most straightforward is to generate an epigenomic map from a cell type or
tissue thought to have an important role in the trait of interest, with the rationale that associated
variants located in regulatory regions are active in the cell type of interest. This approach has
been widely used by several GWASs of face shape and/or disease (25, 99, 163, 164), due in part
to its relative ease: Once epigenomic maps have been generated from sufficient cell numbers in
the relevant ex vivo– or in vitro–derived cell type, they can be reused in future GWASs, and the fine
mapping can be performed genome-wide. Stronger evidence can be provided by perturbing the
activity of candidate regulatory elements harboring associated variants and assessing the impact on
the expression of nearby genes. Yet stronger evidence results from re-creating the precise (typically
single-nucleotide) changes of each trait-associated variant and assessing their impact on regulatory
element activity and target gene expression.
Several studies have used experimental fine-mapping approaches beyond epigenomic maps.
Xiong et al. (170) carried out a GWAS of univariate facial distances from multiple ancestries and
performed both epigenomic fine mapping and luciferase reporter assays of resulting candidate
SNPs in neural crest cells, finding significant effects for four out of five SNPs tested. Two recent
in-depth studies of noncoding mutations underlying craniofacial syndromes provide further exam-
ples of what can be accomplished. Long et al. (97) combined epigenomic maps, enhancer reporter
assays, and endogenous deletion of enhancers in both in vitro–derived CNCCs and mice to show
that structural variants causing Pierre Robin sequence disrupt enhancers more than 1 Mb away
from SOX9. Hirsch et al. (59) used similar approaches to demonstrate that structural variants en-
compassing parts of HDAC9, a neighboring gene to the craniofacial transcription factor TWIST1,
cause craniosynostosis through disruption of TWIST1 enhancers. These studies were in part suc-
cessful at fine mapping the regulatory elements underlying disease-causing mutations because of
their relatively large effect size, but this becomes more challenging for variants associated with
common variation in facial shape or risk for nsCL/P or craniosynostosis.
400 Naqvi et al.

While both Long et al. (97) and Hirsch et al. (59) provided strong evidence for causality by
re-creating variants through genome editing in their endogenous regulatory context, such an
approach is both time and resource intensive, especially if performed in relevant model systems,
such as human pluripotent stem cell–derived CNCCs and their derivatives, or mice. A set of at-
tractive in-between options is provided by massively parallel reporter assays and their numerous
variants, all of which involve transfection or transduction of a library of DNA constructs consisting
of regulatory elements of interest upstream of a reporter gene with genetic barcodes, allowing for
sequencing-based readouts of all regulatory element activity simultaneously [reviewed by Inoue
& Ahituv (69)]. Thus, the effect of multiple candidate variants located in hundreds or thousands of
regulatory elements can be simultaneously assayed. We envision that such an approach could be
applied to the many loci already discovered by facial shape GWASs, and the most prominent hits
could be followed up with more resource-intensive approaches, including re-creating the precise
variants in their endogenous context.

Fundamentally, genetic studies map variants associated with a phenotype; thus, connecting
variants to genes is a key component of understanding phenotypic variation. As with most GWASs
of complex traits and common diseases, most of the lead hits from studies of common facial shape
variation and disease lie in noncoding regions of the genome. Again, the gold standard for con-
necting variants to target genes is re-creating naturally occurring variants endogenously and as-
sessing their impact on nearby target genes, but as with fine mapping, higher-throughput ap-
proaches are necessary for mapping the target genes of the hundreds of variants associated with
facial shape or disease to date. One possible higher-throughput approach is the generation of
molecular QTL data sets, which involve measuring molecular traits such as RNA or protein levels
or splicing in tissue or cells from a population of individuals. Genetic variants in this population
can then be associated with molecular effects on a specific gene, and if such QTLs overlap with
trait-associated variants, the associated genes are candidates for mediating the variant–trait as-
sociation. Although eQTL data sets have yielded some success in fine mapping and target gene
identification for other traits, eQTLs can miss the relevant target genes for a number of reasons,
such as assaying cell types that are not important for the trait or insufficient statistical power
[reviewed by Umans et al. (155)]. Due to the limited availability of primary craniofacial tissue
from the relevant time points (i.e., during embryonic development), a more attainable avenue may
be to use induced pluripotent stem cells (iPSCs) from a population of individuals and differentiate
them into the relevant cell types (i.e., CNCCs or their chondrocyte/osteoblast derivatives). Studies
in iPSC-derived cardiomyocytes and neurons suggest that sample sizes of at least 20–80 individ-
uals are required to detect eQTLs with moderately large effect sizes when performing parallel
differentiations of iPSCs from different individuals combined with bulk RNA sequencing (134,
146). Recent innovative approaches based on pooled differentiations of iPSCs from many indi-
viduals, combined with single-cell RNA sequencing to assign cells to individuals based on genetic
polymorphisms, can both increase power by reducing noise due to differentiation batch effects
and allow for larger sample sizes to be assayed in fewer experiments (32, 71).
A complementary approach to population-based studies is to map the target gene of the reg-
ulatory element (i.e., the enhancer) in which candidate causal variants (identified by some fine-
mapping approach) reside. Broadly, these methods are based on the idea that enhancers come
into close physical proximity to the promoters of their target genes, a feature that can be mapped
genome-wide by chromosome conformation capture assays. Recent work has integrated such as-
says with other epigenomic data sets (i.e., measures of chromatin accessibility and histone modi-
fications) to predict target genes of enhancers (47), with appreciable precision in applications to
GWASs of common diseases, such as inflammatory bowel disease (113). Pairing chromosome con-
formation capture with chromatin immunoprecipitation (HiChIP) of active enhancer marks such

as H3K27ac has also been used to map target genes of SNPs associated with both autoimmune
and cardiovascular disease (110) as well as with prostate cancer (48).
Epigenomic Maps from Relevant Cell Types

As mentioned throughout this review, epigenomic maps from cell types relevant to craniofacial
shape have many uses in understanding variation, both in nominating the cell types and develop-
mental times most important for variation and in fine-mapping variants. Thus far, such data sets
in humans have been generated primarily from in vitro pluripotent stem cell–derived CNCCs or
from bulk tissue samples from a few primary embryonic craniofacial stages. Numerous additional
cell types and developmental times are important for craniofacial development, as evidenced by
mouse studies, and may also explain additional variation in craniofacial morphology. These cell
types include, for example, facial ectoderm, which secretes morphogens such as Shh, important for
patterning CNCCs during early development (41), as well as cell types important for later bone
remodeling, such as osteoclasts and recently identified skeletal stem cells (21). As discussed in the
introduction, cross-species chimeras suggest that CNCCs are the main drivers of species-specific
craniofacial patterning, whereas surface ectoderm is more important for the species-generic pat-
tern (133); such logic may also apply to within-species variation, but this remains to be seen. Ide-
ally, cell types would be assayed at a range of developmental times from primary human samples,
but this is challenging due to both the complex nature of such samples (i.e., containing many cell
types) and their limited availability.
The first challenge can be addressed using single-cell epigenomic profiling technologies, which
currently provide single-cell, genome-wide data at low depth, but when grouped by cell type pro-
vide depth similar to that of bulk assays (139). The overall lack of availability of primary sam-
ples (as well as the inability to use them for further functional tests, for obvious ethical reasons)
is one that may be partially ameliorated by the increasing success of in vitro organoid model-
ing approaches to recapitulate aspects of early human development. For example, a recent study
used micropatterned chips to achieve robust morphogenesis through folding and closure of the
neural tube, starting from commonly used pluripotent stem cells (77). Remarkably, these in vitro–
derived neural tubes gave rise to migratory neural crest cells, which could be expanded and profiled
further.
Mouse Models of Human Craniofacial Variation

While many aspects of human craniofacial development can be probed using the approaches de-
scribed above, they clearly lack the ability to experimentally perturb the development of a true
craniofacial structure in vivo. Thus, despite its evolutionary distance and clear difference in over-
all craniofacial shape from humans, the mouse remains the most accessible and relevant in vivo
model for testing hypotheses arising from studies of human craniofacial variation. Indeed, of the
343 genes associated with a Mendelian craniofacial syndrome in OMIM, 151 (44%) have ortholo-
gous genes in mice that result in craniofacial phenotypes when mutated, as indicated by the Mouse
Phenome Database (Figure 1). Overall, this statistic indicates a high degree of conservation of the
pathways underlying craniofacial development between human and mouse. Furthermore, given
that many of these mutant mouse models were created after gene mapping of specific craniofacial
syndromes, it also further illustrates the success of Mendelian craniofacial genetics in uncovering
important regulators of craniofacial development.
Mouse studies based on GWASs of facial shape variation or nsCL/P risk are more challeng-
ing than those based on the human genetics of rare craniofacial disorders. This is in large part
because most GWAS loci reside in noncoding parts of the genome, which in general are not well
402 Naqvi et al.

conserved between human and mouse. Even when the orthologous noncoding region exists in the
mouse genome, its sequence, regulatory activity, or relative contribution to the target gene expres-
sion is often poorly conserved. For example, in a study by Long et al. (97) of enhancers of SOX9
underlying Pierre Robin sequence, deletion of the orthologous conserved enhancers in mouse
resulted in relatively subtle changes in Sox9 expression and lower jaw morphology, as compared
with the large corresponding effects associated with the loss of human enhancers in CNCCs and
patients. Indeed, a recent study of two genes highlighted specifically by facial shape GWASs, Pax1
and Tbx15, resorted to mutating protein-coding sequence, as the candidate regulatory elements
underlying the GWAS loci are not known with certainty and are in any case not well conserved in
mouse (124). Similarly, Adhikari et al. (1) used gain- and loss-of-function mutants of Edar, SNPs
near which were associated with chin protrusion in a Latin American cohort, to demonstrate an
effect on mandible length.
One potential solution to the lack of noncoding sequence conservation between human and
mouse would be to insert variants of the human regulatory element underlying a GWAS hit (deter-
mined using the experimental fine-mapping approaches discussed above) into the syntenic region
of the mouse genome and then assess the resulting effects on target gene expression and cranio-
facial morphology. However, it is possible, if not likely, that the regulatory landscape surrounding
the mouse orthologous locus has diverged enough that the perturbation may not reflect regulatory
changes equivalent to those influenced by the variant in the context of human development.
FUTURE DIRECTIONS
The human face is a highly variable and multipartite structure resulting from coordinated action
between diverse genetic and nongenetic factors. Recent advances in phenotyping strategies and
analytical approaches have substantially improved our understanding of the genetic contributions
to craniofacial morphology, and we expect this trend to continue in future years, though some
challenges remain.
GWASs have illuminated the polygenic nature of craniofacial morphology, uncovering more
than 300 genetic loci associated with one or multiple aspects of facial form. One of the major
challenges continues to be the identification of the associated variants for function and the mech-
anisms by which they alter gene function or expression. In addition, individual variants contribute
only a small proportion of the phenotypic variance, and a large number of genetic loci are yet to
be discovered, since they likely have effect sizes that are too small or allele frequencies that are too
rare to be detected with statistical significance using current sample sizes, which are still modest by
modern GWAS standards. Furthermore, epistatic effects between combinations of variants acting
on overlapping areas of the face shape space likely also have relatively small effects and thus remain
undiscovered. A natural way to increase statistical power is to increase the size of the study cohort,
for instance, through joint collaborative efforts. For traditional anthropometric measurements,
this is relatively straightforward, but it becomes more challenging when principal components or
multivariate descriptions of the face are used. The latter requires sharing individual raw images or
the underlying principal component analysis constructs, which is often not possible due to strict
privacy considerations.
Combining data sets is further complicated when individuals of diverse ancestral backgrounds
or developmental stages are included. Although the need for increasing sample diversity in genetic
studies is now widely appreciated, most facial shape GWASs have been performed in relatively ho-
mogeneous cohorts of primarily European descent. In cases of population heterogeneity, proper
statistical considerations are required to control for population stratification. One approach is to
stratify samples into major population groups, as determined by principal component analysis, and

combine the results in a cross-ancestry meta-analysis. Alternatively, the joint analysis of the full
cohort offers advantages in terms of statistical power for discovery (168) and its ability to generate
ancestry-specific estimates (2). To deal with age as a confounder, studies have either analyzed in-
dividuals at a single developmental time point or removed the effects of age statistically. However,
such corrections are inherently flawed, as facial growth and development and, later in life, facial
aging are complex, nonlinear processes. Nonlinear kernel-based methods have been developed
for three-dimensional facial data (104, 105), but the collection of longitudinal data will also offer
unique insights into the modeling of craniofacial growth.
Larger cohorts of either healthy individuals or those with craniofacial disorders will also allow
for a further blurring of the distinction between common and rare variation in genetic studies.
For example, sufficiently large cohorts of patients with neurodevelopmental disorders allowed
for the discovery of a common variant burden modulating disorder risk (116); in the context of
craniofacial disorders, such an approach could be used to test whether distinct syndromes have
correlated but not identical common variant burdens, as predicted by the multivariate liability
threshold model. Large cohorts of healthy individuals would allow for the identification of in-
dividuals who have rare mutations in important craniofacial genes but no disease, which could
reveal additional environmental or genetic factors that reduce disease penetrance. Larger sample
sizes would also allow for the explicit modeling of gene–environment interactions, which remain
underexplored with respect to facial variation.
As our knowledge of the genetic architecture of the human face grows, it will become possi-
ble to use this information in concrete, real-life applications. Perhaps one of the most appealing
applications in human facial genetics is the prediction or reconstruction of facial features from bio-
logical material, for example, collected from patients [e.g., to guide orthodontic treatment or facial
reconstructive surgery (72)], recovered from skeletal remains [e.g., to visualize the appearance of
hominin ancestors (49)], or left at a crime scene [e.g., to narrow down potential suspects (23)].
Forensic DNA phenotyping, or the prediction of physical appearance from DNA, shows promise
when comparative DNA profiling fails (i.e., when no match between the unknown sample and
suspect or known profiles in a DNA database can be obtained) (82). The simultaneous prediction
of eye, hair, and skin color has already been validated forensically (20). However, current predic-
tion results for facial morphological traits do not reach high enough accuracy, and commercial
tools that claim to do so lack any scientific validation (166). Furthermore, previous studies have
shown that predictions are driven only by ancestry and sex (24, 92, 135), and complex patterns of
interactions among genes and/or with the environment further complicate matters (55). Methods
to predict facial appearance and infer DNA-related features have also raised concerns regarding
privacy and ethics (78, 83, 156), although the actual risk of reidentification was recently shown to
be smaller than previously claimed (157). The establishment of adequate legislation and regulatory
frameworks for any practice or application will be a necessity for its implementation.
DISCLOSURE STATEMENT
J.W. serves on the scientific boards of CAMP4, Paratus, the Whitehead Institute of the Mas-
sachusetts Institute of Technology, the Searle Scholars Program, and the International Institute
of Molecular and Cell Biology in Warsaw.
ACKNOWLEDGMENTS
J.W. was supported by the Howard Hughes Medical Institute, a Lorry Lokey endowed professor-
ship, and a Stinehart Reed award. S.N. was supported by a Helen Hay Whitney Fellowship. H.H.
and P.C. were supported by the National Institutes of Health (grant R01-DE027023), the KU
404 Naqvi et al.

Leuven research fund (grants BOF-C1, C14/15/081, and C14/20/081), and the research program
of the Research Foundation – Flanders (Belgium) (grant G078518N). J.R.S. and S.M.W. were
supported by the National Institute of Dental and Craniofacial Research (grants U01-DE020078,
R01-DE016148, and R01-DE027023). S.W. and F.W. were supported by the National Institutes
of Health (grant R01-DE027023).
LITERATURE CITED
1. Adhikari K, Fuentes-Guajardo M, Quinto-Sánchez M, Mendoza-Revilla J, Chacón-Duque JC, et al.
2016. A genome-wide association scan implicates DCHS2, RUNX2, GLI3, PAX1 and EDAR in human
facial variation. Nat. Commun. 7:11616
2. Atkinson EG, Maihofer AX, Kanai M, Martin AR, Karczewski KJ, et al. 2021. Tractor uses local ancestry
to enable the inclusion of admixed individuals in GWAS and to boost power. Nat. Genet. 53:195–204
3. Baldi P. 2018. Deep learning in biomedical data science. Annu. Rev. Biomed. Data Sci. 1:181–205
4. Bannister JJ, Crites SR, Aponte JD, Katz DC, Wilms M, et al. 2020. Fully automatic landmarking of
syndromic 3D facial surface scans using 2D images. Sensors 20:3171

5. Barton RA, Venditti C. 2014. Rapid evolution of the cerebellum in humans and other great apes. Curr.
Biol. 24:2440–44
6. Beaty TH, Murray JC, Marazita ML, Munger RG, Ruczinski I, et al. 2010. A genome-wide association
study of cleft lip with and without cleft palate identifies risk variants near MAFB and ABCA4. Nat. Genet.
42:525–29
7. Beaty TH, Taub MA, Scott AF, Murray JC, Marazita ML, et al. 2013. Confirming genes influencing risk
to cleft lip with/without cleft palate in a case-parent trio study. Hum. Genet. 132:771–81
8. Birnbaum S, Ludwig KU, Reutter H, Herms S, Steffens M, et al. 2009. Key susceptibility locus for
nonsyndromic cleft lip with or without cleft palate on chromosome 8q24. Nat. Genet. 41:473–77
9. Boehringer S, van der Lijn F, Liu F, Günther M, Sinigerova S, et al. 2011. Genetic determination of
human facial morphology: links between cleft-lips and normal variation. Eur. J. Hum. Genet. 19:1192–97
10. Böhringer S, de Jong MA. 2019. Quantification of facial traits. Front. Genet. 10:397
11. Bonfante B, Faux P, Navarro N, Mendoza-Revilla J, Dubied M, et al. 2021. A GWAS in Latin
Americans identifies novel face shape loci, implicating VPS13B and a Denisovan introgressed region
in facial variation. Sci. Adv. 7:eabc6160
12. Boulet SL, Rasmussen SA, Honein MA. 2008. A population-based study of craniosynostosis in
metropolitan Atlanta, 1989–2003. Am. J. Med. Genet. A 146A:984–91
13. Bronstein MM, Bruna J, LeCun Y, Szlam A, Vandergheynst P. 2017. Geometric deep learning: going
beyond Euclidean data. IEEE Signal Process. Mag. 34(4):18–42
14. Butaric LN, Klocke RP. 2018. Nasal variation in relation to high-altitude adaptations among Tibetans
and Andeans. Am. J. Hum. Biol. 30:e23104
15. Calo E, Gu B, Bowen ME, Aryan F, Zalc A, et al. 2018. Tissue-selective effects of nucleolar stress and
rDNA damage in developmental disorders. Nature 554:112–17
16. Carlson JC, Standley J, Petrin A, Shaffer JR, Butali A, et al. 2017. Identification of 16q21 as a modifier
of nonsyndromic orofacial cleft phenotypes. Genet. Epidemiol. 41:887–97
17. Casey BJ, Cannonier T, Conley MI, Cohen AO, Barch DM, et al. 2018. The Adolescent Brain Cognitive
Development (ABCD) study: imaging acquisition across 21 sites. Dev. Cogn. Neurosci. 32:43–54
18. Castel SE, Cervera A, Mohammadi P, Aguet F, Reverter F, et al. 2018. Modified penetrance of coding
variants by cis-regulatory variation contributes to disease risk. Nat. Genet. 50:1327–34
19. Cha S, Lim JE, Park AY, Do J-H, Lee SW, et al. 2018. Identification of five novel genetic loci related to
facial morphology by genome-wide association studies. BMC Genom. 19:481
20. Chaitanya L, Breslin K, Zuñiga S, Wirken L, Pośpiech E, et al. 2018. The HIrisPlex-S system for eye,
hair and skin colour prediction from DNA: introduction and forensic developmental validation. Forensic
Sci. Int. Genet. 35:123–35
21. Chan CKF, Gulati GS, Sinha R, Tompkins JV, Lopez M, et al. 2018. Identification of the human skeletal
stem cell. Cell 175:43–56.e21

22. Chan CKF, Seo EY, Chen JY, Lo D, McArdle A, et al. 2015. Identification and specification of the mouse
skeletal stem cell. Cell 160:285–98
23. Claes P, Hill H, Shriver MD. 2014. Toward DNA-based facial composites: preliminary results and
validation. Forensic Sci. Int. Genet. 13:208–16
24. Claes P, Liberton DK, Daniels K, Rosana KM, Quillen EE, et al. 2014. Modeling 3D facial shape from
DNA. PLOS Genet. 10:e1004224
25. Claes P, Roosenboom J, White JD, Swigut T, Sero D, et al. 2018. Genome-wide mapping of global-to-
local genetic effects on human facial shape. Nat. Genet. 50:414–23
26. Claes P, Walters M, Clement J. 2012. Improved facial outcome assessment using a 3D anthropometric
mask. Int. J. Oral Maxillofac. Surg. 41:324–30
27. Cole JB, Manyama M, Kimwaga E, Mathayo J, Larson JR, et al. 2016. Genomewide association study
of African children identifies association of SCHIP1 and PDE8A with facial size and shape. PLOS Genet.
12:e1006174
28. Cole JB, Manyama M, Larson JR, Liberton DK, Ferrara TM, et al. 2017. Human facial shape and size
heritability and genetic correlations. Genetics 205:967–78
29. Cordero DR, Brugmann S, Chu Y, Bajpai R, Jame M, et al. 2011. Cranial neural crest cells on the move:
their roles in craniofacial development. Am. J. Med. Genet. A 155A:270–79

30. Crouch DJM, Winney B, Koppen WP, Christmas WJ, Hutnik K, et al. 2018. Genetics of the human
face: identification of large-effect single gene variants. PNAS 115:E676–85
31. Crouzon O. 1912. La dysostose cranio-faciale héréditaire. Bull. Mém. Soc. Med. Hôp. Paris 33:545–55
32. Cuomo ASE, Seaton DD, McCarthy DJ, Martinez I, Bonder MJ, et al. 2020. Single-cell RNA-
sequencing of differentiating iPS cells reveals dynamic genetic effects on gene expression. Nat. Commun.
11:810
33. Curtis SW, Chang D, Lee MK, Shaffer JR, Indencleef K, et al. 2021. The PAX1 locus at 20p11 is a
potential genetic modifier for bilateral cleft lip. Hum. Genet. Genom. Adv. 2:100025
34. Curtis SW, Chang D, Sun MR, Epstein MP, Murray JC, et al. 2021. FAT4 identified as a potential
modifier of orofacial cleft laterality. Genet. Epidemiol. 45:721–35
35. Dardani C, Howe LJ, Mukhopadhyay N, Stergiakouli E, Wren Y, et al. 2020. Cleft lip/palate and educa-
tional attainment: cause, consequence or correlation? A Mendelian randomization study. Int. J. Epidemiol.
49:1282–93
36. de Jong MA, Wollstein A, Ruff C, Dunaway D, Hysi P, et al. 2016. An automatic 3D facial landmarking
algorithm using 2D Gabor wavelets. IEEE Trans. Image Process. 25:580–88
37. Debusmann P. 1940. Familiare kombinierte Gesichtsmissbildung im Bereich des ersten Viszeralbogens.
Arch. Kinderheil. 120:133–39
38. Dixon J, Edwards SJ, Gladwin AJ, Dixon MJ, Loftus SK, et al. 1996. Positional cloning of a gene involved
in the pathogenesis of Treacher Collins syndrome. Nat. Genet. 12:130–36
39. Djordjevic J, Zhurov AI, Richmond S, Visigen Consort. 2016. Genetic and environmental contributions
to facial morphological variation: a 3D population-based twin study. PLOS ONE 11:e0162250. Correc-
tion. 2016. PLOS ONE 11:e0164961
40. Dudas M, Kaartinen V. 2005. Tgf-β superfamily and mouse craniofacial development: interplay of
morphogenetic proteins and receptor signaling controls normal formation of the face. Curr. Top. Dev.
Biol. 66:65–133
41. Dworkin S, Boglev Y, Owens H, Goldie SJ. 2016. The role of Sonic hedgehog in craniofacial patterning,
morphogenesis and cranial neural crest survival. J. Dev. Biol. 4:24
42. Endo C, Johnson TA, Morino R, Nakazono K, Kamitsuji S, et al. 2018. Genome-wide association study
in Japanese females identifies fifteen novel skin-related trait associations. Sci. Rep. 8:8974
43. Enlow DH, Hans MG. 1996. Essentials of Facial Growth. Philadelphia: Saunders
44. Falconer DS. 1996. Introduction to Quantitative Genetics. Harlow, UK: Prentice Hall
45. Finucane HK, Bulik-Sullivan B, Gusev A, Trynka G, Reshef Y, et al. 2015. Partitioning heritability by
functional annotation using genome-wide association summary statistics. Nat. Genet. 47:1228–35
46. Fuad MTH, Fime AA, Sikder D, Iftee MAR, Rabbi J, et al. 2021. Recent advances in deep learning
techniques for face recognition. IEEE Access 9:99112–42
406 Naqvi et al.

47. Fulco CP, Nasser J, Jones TR, Munson G, Bergman DT, et al. 2019. Activity-by-contact model of
enhancer-promoter regulation from thousands of CRISPR perturbations. Nat. Genet. 51:1664–69
48. Giambartolomei C, Seo J-H, Schwarz T, Freund MK, Johnson RD, et al. 2021. H3K27ac HiChIP in
prostate cell lines identifies risk genes for prostate cancer susceptibility. Am. J. Hum. Genet. 108:2284–
300
49. Gokhman D, Mishol N, de Manuel M, de Juan D, Shuqrun J, et al. 2019. Reconstructing Denisovan
anatomy using DNA methylation maps. Cell 179:180–92.e10
50. Graf D, Malik Z, Hayano S, Mishina Y. 2016. Common mechanisms in development and disease: BMP
signaling in craniofacial development. Cytokine Growth Factor Rev. 27:129–39
51. Grant SFA, Wang K, Zhang H, Glaberson W, Annaiah K, et al. 2009. A genome-wide association study
identifies a locus for nonsyndromic cleft lip with or without cleft palate on 8q24. J. Pediatr. 155:909–13
52. Greenberg RS, Long HK, Swigut T, Wysocka J. 2019. Single amino acid change underlies distinct roles
of H2A.Z subtypes in human syndrome. Cell 178:1421–36.e24
53. Guo J, Tan J, Yang Y, Zhou H, Hu S, et al. 2014. Variation and signatures of selection on the human face.
J. Hum. Evol. 75:143–52
54. Gurovich Y, Hanani Y, Bar O, Nadav G, Fleischer N, et al. 2019. Identifying facial phenotypes of genetic
disorders using deep learning. Nat. Med. 25:60–64

55. Hallgrimsson B, Mio W, Marcucio RS, Spritz R. 2014. Let’s face it—complex traits are just not that
simple. PLOS Genet. 10:e1004724
56. Hallgrimsson B, Percival CJ, Green R, Young NM, Mio W, et al. 2015. Morphometrics, 3D imaging,
and craniofacial development. In Craniofacial Development, ed. Y Chai, pp. 561–97. Curr. Top. Dev. Biol.
115. Waltham, MA: Academic
57. Hammond P. 2007. The use of 3D face shape modelling in dysmorphology. Arch. Dis. Child. 92:1120–26
58. Hamosh A, Scott AF, Amberger JS, Bocchini CA, McKusick VA. 2005. Online Mendelian Inheritance
in Man (OMIM), a knowledgebase of human genes and genetic disorders. Nucleic Acids Res. 33:D514–17
59. Hirsch N, Dahan I, D’haene E, Avni M, Vergult S, et al. 2021. HDAC9 structural variants dis-
rupting TWIST1 transcriptional regulation lead to craniofacial and limb malformations. bioRxiv
2021.08.10.455254. https://doi.org/10.1101/2021.08.10.455254
60. Hoskens H, Li J, Indencleef K, Gors D, Larmuseau MHD, et al. 2018. Spatially dense 3D facial
heritability and modules of co-heritability in a father-offspring design. Front. Genet. 9:554
61. Hoskens H, Liu D, Naqvi S, Lee MK, Eller RJ, et al. 2021. 3D facial phenotyping by biometric sibling
matching used in contemporary genomic methodologies. PLOS Genet. 17:e1009528
62. Howe LJ, Lee MK, Sharp GC, Davey Smith G, St. Pourcain B, et al. 2018. Investigating the shared
genetics of non-syndromic cleft lip/palate and facial morphology. PLOS Genet. 14:e1007501
63. Hrdlička A. 1920. Anthropometry. Philadelphia: Wistar Inst. Anat. Biol.
64. Huang L, Jia Z, Shi Y, Du Q, Shi J, et al. 2019. Genetic factors define CPO and CLO subtypes of
nonsyndromicorofacial cleft. PLOS Genet. 15:e1008357
65. Huang Y, Li D, Qiao L, Liu Y, Peng Q, et al. 2021. A genome-wide association study of facial morphol-
ogy identifies novel genetic loci in Han Chinese. J. Genet. Genom. 48:198–207
66. Hutton TJ, Buxton BR, Hammond P. 2001. Dense surface point distribution models of the human face.
In Proceedings of the IEEE Workshop on Mathematical Methods in Biomedical Image Analysis (MMBIA 2001),
pp. 153–60. Piscataway, NJ: IEEE
67. Indencleef K, Hoskens H, Lee MK, White JD, Liu C, et al. 2021. The intersection of the genetic
architectures of orofacial clefts and normal facial variation. Front. Genet. 12:170
68. Indencleef K, Roosenboom J, Hoskens H, White JD, Shriver MD, et al. 2018. Six NSCL/P loci show
associations with normal-range craniofacial variation. Front. Genet. 9:502
69. Inoue F, Ahituv N. 2015. Decoding enhancers using massively parallel reporter assays. Genomics
106:159–64
70. Jacobs LC, Liu F, Bleyen I, Gunn DA, Hofman A, et al. 2014. Intrinsic and extrinsic risk factors for
sagging eyelids. JAMA Dermatol. 150:836–43
71. Jerber J, Seaton DD, Cuomo ASE, Kumasaka N, Haldane J, et al. 2021. Population-scale single-cell
RNA-seq profiling across dopaminergic neuron differentiation. Nat. Genet. 53:304–12

72. Jheon AH, Oberoi S, Solem RC, Kapila S. 2017. Moving towards precision orthodontics: an evolving
paradigm shift in the planning and delivery of customized orthodontic therapy. Orthod. Craniofac. Res.
20(Suppl. 1):106–13
73. Johannsdottir B, Thorarinsson F, Thordarson A, Magnusson TE. 2005. Heritability of craniofacial char-
acteristics between parents and offspring estimated from lateral cephalograms. Am. J. Orthod. Dentofac.
Orthop. 127:200–61
74. Jones MC. 1988. Etiology of facial clefts: prospective evaluation of 428 patients. Cleft Palate J. 25:16–20
75. Justice CM, Cuellar A, Bala K, Sabourin JA, Cunningham ML, et al. 2020. A genome-wide association
study implicates the BMP7 locus as a risk factor for nonsyndromic metopic craniosynostosis. Hum. Genet.
139:1077–90
76. Justice CM, Yagnik G, Kim Y, Peter I, Jabs EW, et al. 2012. A genome-wide association study identifies
susceptibility loci for nonsyndromic sagittal craniosynostosis near BMP2 and within BBS9. Nat. Genet.
44:1360–64
77. Karzbrun E, Khankhel AH, Megale HC, Glasauer SMK, Wyle Y, et al. 2021. Human neural tube
morphogenesis in vitro by geometric constraints. Nature 599:268–72
78. Katsanis SH, Claes P, Doerr M, Cook-Deegan R, Tenenbaum JD, et al. 2021. A survey of U.S. public
perspectives on facial recognition technology and facial imaging data practices in health and research
contexts. PLOS ONE 16:e0257923
79. Katz DC, Grote MN, Weaver TD. 2017. Changes in human skull morphology across the agricultural
transition are consistent with softer diets in preindustrial farming groups. PNAS 114:9050–55
80. Kau CH, Richmond S. 2008. Three-dimensional analysis of facial morphology surface changes in
untreated children from 12 to 14 years of age. Am. J. Orthod. Dentofac. Orthop. 134:751–60
81. Kau CH, Richmond S, Zhurov A, Ovsenik M, Tawfik W, et al. 2010. Use of 3-dimensional surface acqui-
sition to study facial morphology in 5 populations. Am. J. Orthod. Dentofac. Orthop. 137(Suppl.):S56.e1–9
82. Kayser M. 2015. Forensic DNA phenotyping: predicting human appearance from crime scene material
for investigative purposes. Forensic Sci. Int. Genet. 18:33–48
83. Kayser M, de Knijff P. 2011. Improving human forensics through advances in genetics, genomics and
molecular biology. Nat. Rev. Genet. 12:179–92
84. Kesterke MJ, Raffensperger ZD, Heike CL, Cunningham ML, Hecht JT, et al. 2016. Using the 3D
Facial Norms Database to investigate craniofacial sexual dimorphism in healthy children, adolescents,
and adults. Biol. Sex Differ. 7:23
85. Knol MJ, Pawlak MA, Lamballais S, Terzikhan N, Hofer E, et al. 2022. Genetic architecture of orbital
telorism. Hum. Mol. Genet. 31:1531–43
86. Kohn LAP. 1991. The role of genetics in craniofacial morphology and growth. Annu. Rev. Anthropol.
20:261–78
87. Lajeunie E, Le Merrer M, Bonaïti-Pellie C, Marchac D, Renier D. 1995. Genetic study of nonsyndromic
coronal craniosynostosis. Am. J. Med. Genet. 55:500–4
88. Lee MK, Shaffer JR, Leslie EJ, Orlova E, Carlson JC, et al. 2017. Genome-wide association study of
facial morphology reveals novel associations with FREM1 and PARK2. PLOS ONE 12:e0176566
89. Leslie EJ, Carlson JC, Shaffer JR, Butali A, Buxó CJ, et al. 2017. Genome-wide meta-analyses of non-
syndromic orofacial clefts identify novel associations between FOXE1 and all orofacial clefts, and TP63
and cleft lip with or without cleft palate. Hum. Genet. 136:275–86
90. Li M, Cole JB, Manyama M, Larson JR, Liberton DK, et al. 2017. Rapid automated landmarking for
morphometric analysis of three-dimensional facial scans. J. Anat. 230:607–18
91. Lieberman P. 2007. The evolution of human speech: its anatomical and neural bases. Curr. Anthropol.
48:39–66
92. Lippert C, Sabatini R, Maher MC, Kang EY, Lee S, et al. 2017. Identification of individuals by trait
prediction using whole-genome sequencing data. PNAS 114:10166–71
93. Liu C, Lee MK, Naqvi S, Hoskens H, Liu D, et al. 2021. Genome scans of facial features in East Africans
and cross-population comparisons reveal novel associations. PLOS Genet. 17:e1009695
94. Liu D, Alhazmi N, Matthews H, Lee MK, Li J, et al. 2021. Impact of low-frequency coding variants on
human facial shape. Sci. Rep. 11:748
408 Naqvi et al.

95. Liu D, Ban H-J, El Sergani AM, Lee MK, Hecht JT, et al. 2021. PRICKLE1 × FOCAD interaction
revealed by genome-wide vQTL analysis of human facial traits. Front. Genet. 12:674642
96. Liu F, van der Lijn F, Schurmann C, Zhu G, Chakravarty MM, et al. 2012. A genome-wide association
study identifies five loci influencing facial morphology in Europeans. PLOS Genet. 8:e1002932
97. Long HK, Osterwalder M, Welsh IC, Hansen K, Davies JOJ, et al. 2020. Loss of extreme long-range
enhancers in human neural crest drives a craniofacial disorder. Cell Stem Cell 27:765–83.e14
98. Ludwig KU, Ahmed ST, Böhmer AC, Sangani NB, Varghese S, et al. 2016. Meta-analysis reveals
genome-wide significance at 15q13 for nonsyndromic clefting of both the lip and the palate, and func-
tional analyses implicate GREM1 as a plausible causative gene. PLOS Genet. 12:e1005914
99. Ludwig KU, Böhmer AC, Bowes J, Nikolic M, Ishorst N, et al. 2017. Imputation of orofacial clefting
data identifies novel risk loci and sheds light on the genetic background of cleft lip ± cleft palate and
cleft palate only. Hum. Mol. Genet. 26:829–42
100. Ludwig KU, Mangold E, Herms S, Nowak S, Reutter H, et al. 2012. Genome-wide meta-analyses of
nonsyndromic cleft lip with or without cleft palate identify six new risk loci. Nat. Genet. 44:968–71
101. Mahdi SS, Nauwelaers N, Joris P, Bouritsas G, Gong S, et al. 2022. Matching 3D facial shape to
demographic properties by geometric metric learning: a part-based approach. IEEE Trans. Biometr.
Behav. Identity Sci. 4:163–72

102. Mangold E, Ludwig KU, Birnbaum S, Baluardo C, Ferrian M, et al. 2010. Genome-wide association
study identifies two susceptibility loci for nonsyndromic cleft lip with or without cleft palate. Nat. Genet.
42:24–26
103. Marazita ML, Lidral AC, Murray JC, Field LL, Maher BS, et al. 2009. Genome scan, fine-mapping, and
candidate gene analysis of non-syndromic cleft lip with or without cleft palate reveals phenotype-specific
differences in linkage and association results. Hum. Hered. 68:151–70
104. Matthews HS, Palmer RL, Baynam GS, Quarrell OW, Klein OD, et al. 2021. Large-scale open-
source three-dimensional growth curves for clinical facial assessment and objective description of facial
dysmorphism. Sci. Rep. 11:12175
105. Matthews HS, Penington AJ, Hardiman R, Fan Y, Clement JG, et al. 2018. Modelling 3D craniofacial
growth trajectories for population comparison and classification illustrated using sex-differences. Sci.
Rep. 8:4771
106. Mitteroecker P, Gunz P, Bernhard M, Schaefer K, Bookstein FL. 2004. Comparison of cranial ontoge-
netic trajectories among great apes and humans. J. Hum. Evol. 46:679–97
107. Mostowska A, Gaczkowska A, Żukowski K, Ludwig KU, Hozyasz KK, et al. 2018. Common variants in
DLG1 locus are associated with non-syndromic cleft lip with or without cleft palate. Clin. Genet. 93:784–
93
108. Muggli E, Matthews H, Penington A, Claes P, O’Leary C, et al. 2017. Association between prenatal
alcohol exposure and craniofacial shape of children at 12 months of age. JAMA Pediatr. 171:771–80
109. Mukhopadhyay N, Feingold E, Moreno-Uribe L, Wehby G, Valencia-Ramirez LC, et al. 2021. Genome-
wide association study of non-syndromic orofacial clefts in a multiethnic sample of families and controls
identifies novel regions. Front. Cell Dev. Biol. 9:621482
110. Mumbach MR, Satpathy AT, Boyle EA, Dai C, Gowen BG, et al. 2017. Enhancer connectome in primary
human cells identifies target genes of disease-associated DNA elements. Nat. Genet. 49:1602–12
111. Naini FB, Moss JP. 2004. Three-dimensional assessment of the relative contribution of genetics and
environment to various facial parameters with the twin method. Am. J. Orthod. Dentofac. Orthop. 126:655–
65
112. Naqvi S, Sleyp Y, Hoskens H, Indencleef K, Spence JP, et al. 2021. Shared heritability of human face
and brain shape. Nat. Genet. 53:830–39
113. Nasser J, Bergman DT, Fulco CP, Guckelberger P, Doughty BR, et al. 2021. Genome-wide enhancer
maps link risk variants to disease genes. Nature 593:238–43
114. Neubauer S, Hublin J-J, Gunz P. 2018. The evolution of modern human brain shape. Sci. Adv. 4:eaao5961
115. Nie X, Luukko K, Kettunen P. 2006. FGF signalling in craniofacial development and developmental
disorders. Oral Dis. 12:102–11
116. Niemi MEK, Martin HC, Rice DL, Gallone G, Gordon S, et al. 2018. Common genetic variants
contribute to risk of rare severe neurodevelopmental disorders. Nature 562:268–71

117. Parker SE, Mai CT, Canfield MA, Rickard R, Wang Y, et al. 2010. Updated national birth prevalence
estimates for selected birth defects in the United States, 2004–2006. Birth Defects Res. A 88:1008–16
118. Paternoster L, Zhurov AI, Toma AM, Kemp JP, St. Pourcain B, et al. 2012. Genome-wide associa-
tion study of three-dimensional facial morphology identifies a variant in PAX3 associated with nasion
position. Am. J. Hum. Genet. 90:478–85
119. Pecora NG, Baccetti T, McNamara JA. 2008. The aging craniofacial complex: a longitudinal cephalo-
metric study from late adolescence to late adulthood. Am. J. Orthod. Dentofac. Orthop. 134:496–505
120. Petrides G, Clark JR, Low H, Lovell N, Eviston TJ. 2021. Three-dimensional scanners for soft-tissue
facial assessment in clinical practice. J. Plast. Reconstr. Aesthet. Surg. 74:605–14
121. Pickrell JK, Berisa T, Liu JZ, Ségurel L, Tung JY, et al. 2016. Detection and interpretation of shared
genetic influences on 42 human traits. Nat. Genet. 48:709–17
122. Prescott SL, Srinivasan R, Marchetto MC, Grishina I, Narvaiza I, et al. 2015. Enhancer divergence and
cis-regulatory evolution in the human and chimp neural crest. Cell 163:68–83
123. Proffit W, Fields H, Larson B, Sarver D. 2018. Contemporary Orthodontics. Philadelphia: Elsevier. 6th ed.
124. Qian Y, Xiong Z, Li Y, Zhou H, Kayser M, et al. 2021. Knocking-out the human face genes TBX15 and
PAX1 in mice alters facial and other physical morphology. bioRxiv 2021.05.26.445773. https://doi.org/
10.1101/2021.05.26.445773
125. Qiao L, Yang Y, Fu P, Hu S, Zhou H, et al. 2018. Genome-wide variants of Eurasian facial shape differ-
entiation and a prospective model of DNA based face prediction. J. Genet. Genom. 45:419–32
126. Ransom RC, Carter AC, Salhotra A, Leavitt T, Marecic O, et al. 2018. Mechanoresponsive stem cells
acquire neural crest fate in jaw regeneration. Nature 563:514–21
127. Reardon W, Winter RM, Rutland P, Pulleyn LJ, Jones BM, et al. 1994. Mutations in the fibroblast
growth factor receptor 2 gene cause Crouzon syndrome. Nat. Genet. 8:98–103
128. Reynolds K, Kumari P, Sepulveda Rincon L, Gu R, Ji Y, et al. 2019. Wnt signaling in orofacial clefts:
crosstalk, pathogenesis and models. Dis. Models Mech. 12:dmm037051
129. Ritchie MD, Van Steen K. 2018. The search for gene-gene interactions in genome-wide association
studies: challenges in abundance of methods, practical considerations, and biological interpretation. Ann.
Transl. Med. 6:157
130. Roosenboom J, Indencleef K, Lee MK, Hoskens H, White JD, et al. 2018. SNPs associated with
testosterone levels influence human facial morphology. Front. Genet. 9:497
131. Rudy HL, Wake N, Yee J, Garfein ES, Tepper OM. 2020. Three-dimensional facial scanning at the fin-
gertips of patients and surgeons: accuracy and precision testing of iPhone X three-dimensional scanner.
Plast. Reconstr. Surg. 146:1407–17
132. Schaid DJ, Chen W, Larson NB. 2018. From genome-wide associations to candidate causal variants by
statistical fine-mapping. Nat. Rev. Genet. 19:491–504
133. Schneider RA. 2018. Neural crest and the origin of species-specific pattern. Genesis 56:e23219
134. Schwartzentruber J, Foskolou S, Kilpinen H, Rodrigues J, Alasoo K, et al. 2018. Molecular and functional
variation in iPSC-derived sensory neurons. Nat. Genet. 50:54–61
135. Sero D, Zaidi A, Li J, White JD, Zarzar TBG, et al. 2019. Facial recognition from DNA using face-to-
DNA classifiers. Nat. Commun. 10:2557
136. Sforza C, de Menezes M, Ferrario V. 2013. Soft- and hard-tissue facial anthropometry in three dimen-
sions: what’s new. J. Anthropol. Sci. 91:159–84
137. Shaffer JR, Orlova E, Lee MK, Leslie EJ, Raffensperger ZD, et al. 2016. Genome-wide association study
reveals multiple loci influencing normal human facial morphology. PLOS Genet. 12:e1006149
138. Sheehan MJ, Nachman MW. 2014. Morphological and population genomic evidence that human faces
have evolved to signal individual identity. Nat. Commun. 5:4800
139. Shema E, Bernstein BE, Buenrostro JD. 2019. Single-cell and single-molecule epigenomics to uncover
genome regulation at unprecedented resolution. Nat. Genet. 51:19–25
140. Shin JH, Persing JA. 2007. Nonsyndromic craniosynostosis and deformational plagiocephaly. In Grabb
and Smith’s Plastic Surgery, ed. CH Thorne, RW Beasley, SJ Aston, SP Bartlett, GC Gurtner, SL Spear,
pp. 226–36. Philadelphia: Lippincott Williams & Wilkins. 6th ed.
410 Naqvi et al.

141. Shrimpton S, Daniels K, de Greef S, Tilotta F, Willems G, et al. 2014. A spatially-dense regression study
of facial form and tissue depth: towards an interactive tool for craniofacial reconstruction. Forensic Sci.
Int. 234:103–10
142. Simões-Costa M, Bronner ME. 2015. Establishing neural crest identity: a gene regulatory recipe.
Development 142:242–57
143. Smith DW. 1982. Recognizable Patterns of Human Malformation: Genetic, Embryologic and Clinical Aspects.
Philadelphia: Saunders. 3rd ed.
144. Som PM, Naidich TP. 2013. Illustrated review of the embryology and development of the facial region,
part 1: early face and lateral nasal cavities. Am. J. Neuroradiol. 34:2233–40
145. Som PM, Streit A, Naidich TP. 2014. Illustrated review of the embryology and development of the
facial region, part 3: an overview of the molecular interactions responsible for facial development. Am.
J. Neuroradiol. 35:223–29
146. Strober BJ, Elorbany R, Rhodes K, Krishnan N, Tayeb K, et al. 2019. Dynamic genetic regulation of
gene expression during cellular differentiation. Science 364:1287–90

147. Subtelny JD. 1959. A longitudinal study of soft tissue facial structures and their profile characteristics,
defined in relation to underlying skeletal structures. Am. J. Orthod. 45:481–507

148. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, et al. 2015. UK Biobank: an open access resource for
identifying the causes of a wide range of complex diseases of middle and old age. PLOS Med. 12:e1001779
149. Sun Y, Huang Y, Yin A, Pan Y, Wang Y, et al. 2015. Genome-wide association study identifies a new
susceptibility locus for cleft lip with or without a cleft palate. Nat. Commun. 6:6414
150. Suttie M, Foroud T, Wetherill L, Jacobson JL, Molteno CD, et al. 2013. Facial dysmorphism across the
fetal alcohol spectrum. Pediatrics 131:e779–88
151. Thayer ZM, Dobson SD. 2013. Geographic variation in chin shape challenges the universal facial
attractiveness hypothesis. PLOS ONE 8:e60681
152. Thieme F, Ludwig KU. 2017. The role of noncoding genetic variation in isolated orofacial clefts.
J. Dent. Res. 96:1238–47
153. Timberlake AT, Choi J, Zaidi S, Lu Q, Nelson-Williams C, et al. 2016. Two locus inheritance of
non-syndromic midline craniosynostosis via rare SMAD6 and common BMP2 alleles. eLife 5:e20125
154. Tsagkrasoulis D, Hysi P, Spector T, Montana G. 2017. Heritability maps of human face morphology
through large-scale automated three-dimensional phenotyping. Sci. Rep. 7:45885
155. Umans BD, Battle A, Gilad Y. 2021. Where are the disease-associated eQTLs? Trends Genet. 37:109–24
156. Van Noorden R. 2020. The ethical questions that haunt facial-recognition research. Nature 587:354–58
157. Venkatesaramani R, Malin BA, Vorobeychik Y. 2021. Re-identification of individuals in genomic datasets
using public face images. Sci. Adv. 7:eabg3296
158. Visscher PM, Hill WG, Wray NR. 2008. Heritability in the genomics era—concepts and misconcep-
tions. Nat. Rev. Genet. 9:255–66
159. Vortkamp A, Gessler M, Grzeschik K-H. 1991. GLI3 zinc-finger gene interrupted by translocations in
Greig syndrome families. Nature 352:539–40
160. Wagner T, Wirth J, Meyer J, Zabel B, Held M, et al. 1994. Autosomal sex reversal and campomelic
dysplasia are caused by mutations in and around the SRY-related gene SOX9. Cell 79:1111–20
161. Wang S, Zhang M, Wu S, Du S, Qian W, et al. 2021. Genetic mechanisms underlying East Asian and
European facial differentiation. Res. Square 604881. https://doi.org/10.21203/rs.3.rs-604881/v1
162. Weinberg SM, Parsons TE, Marazita ML, Maher BS. 2013. Heritability of face shape in twins: a
preliminary study using 3D stereophotogrammetry and geometric morphometrics. Dent. 3000 1:14
163. Welzenbach J, Hammond NL, Nikolić M, Thieme F, Ishorst N, et al. 2021. Integrative approaches
generate insights into the architecture of non-syndromic cleft lip with or without cleft palate. Hum.
Genet. Genom. Adv. 2:100038
164. White JD, Indencleef K, Naqvi S, Eller RJ, Hoskens H, et al. 2021. Insights into the genetic architecture
of the human face. Nat. Genet. 53:45–53
165. White JD, Ortega-Castrillon A, Virgo C, Indencleef K, Hoskens H, et al. 2020. Sources of variation in
the 3dMDface and Vectra H1 3D facial imaging systems. Sci. Rep. 10:4443

166. Wienroth M. 2020. Socio-technical disagreements as ethical fora: Parabon NanoLab’s forensic DNA
SnapshotTM service at the intersection of discourses around robust science, technology validation, and
commerce. Biosocieties 15:28–45
167. Wilderman A, VanOudenhove J, Kron J, Noonan JP, Cotney J. 2018. High-resolution epigenomic atlas
of human embryonic craniofacial development. Cell Rep. 23:1581–97
168. Wojcik GL, Graff M, Nishimura KK, Tao R, Haessler J, et al. 2019. Genetic analyses of diverse popula-
tions improves discovery for complex traits. Nature 570:514–18
169. Wu W, Zhai G, Xu Z, Hou B, Liu D, et al. 2019. Whole-exome sequencing identified four loci
influencing craniofacial morphology in northern Han Chinese. Hum. Genet. 138:601–11
170. Xiong Z, Dankova G, Howe LJ, Lee MK, Hysi PG, et al. 2019. Novel genetic loci affecting facial shape
variation in humans. eLife 8:e49898
171. Yang J, Lee SH, Goddard ME, Visscher PM. 2011. GCTA: a tool for genome-wide complex trait analysis.
Am. J. Hum. Genet. 88:76–82
172. Yengo L, Sidorenko J, Kemper KE, Zheng Z, Wood AR, et al. 2018. Meta-analysis of genome-wide
association studies for height and body mass index in ∼700 000 individuals of European ancestry. Hum.
Mol. Genet. 27:3641–49
173. Yu Y, Zuo X, He M, Gao J, Fu Y, et al. 2017. Genome-wide analyses of non-syndromic cleft lip with
palate identify 14 novel loci and genetic heterogeneity. Nat. Commun. 8:14364
174. Zaidi AA, Mattern BC, Claes P, McEcoy B, Hughes C, et al. 2017. Investigating the case of human nose
shape and climate adaptation. PLOS Genet. 13:e1006616
175. Zelditch ML, Swiderski DL, Sheets HD, Fink WL. 2004. Geometric Morphometrics for Biologists: A Primer.
Boston: Academic
412 Naqvi et al.

GG23_TOC ARjats.cls June 16, 2022 14:16
Annual Review of
Genomics and
Contents Human Genetics
Volume 23, 2022
A Journey Through Genetics to Biology

Veronica van Heyningen p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 1
Extrachromosomal DNA in Cancer
Vineet Bafna and Paul S. Mischel p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p29

Maintaining Transcriptional Specificity Through Mitosis
Kenji Ito and Kenneth S. Zaret p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p53

Natural and Experimental Rewiring of Gene Regulatory Regions
Damien J. Downes and Jim R. Hughes p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p73
Obtaining Complete Human Proteomes
Ana Martinez-Val, Ulises H. Guzmán, and Jesper V. Olsen p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p99
Structural Variation in Cancer: Role, Prevalence, and Mechanisms
Marco Raffaele Cosenza, Bernardo Rodriguez-Martin, and Jan O. Korbel p p p p p p p p p p p p p p 123
Genome-Wide Analysis of Human Long Noncoding RNAs:
A Provocative Review
Chris P. Ponting and Wilfried Haerty p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 153
Establishing the Medical Actionability of Genomic Variants
Katrina A.B. Goddard, Kristy Lee, Adam H. Buchanan, Bradford C. Powell,
and Jessica Ezzell Hunter p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 173
Genetic Disorders of the Extracellular Matrix: From Cell and Gene
Therapy to Future Applications in Regenerative Medicine
Shukti Chakravarti, Elena Enzo, Maithê Rocha Monteiro de Barros,
Maria Benedetta Rizzarda Maffezzoni, and Graziella Pellegrini p p p p p p p p p p p p p p p p p p p p p 193
The Genetics and Typical Traits of Thoracic Aortic Aneurysm
and Dissection
Jotte Rodrigues Bento, Josephina Meester, Ilse Luyckx, Silke Peeters,
Aline Verstraeten, and Bart Loeys p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 223
The Genetics of Brugada Syndrome
Marina Cerrone, Sarah Costa, and Mario Delmar p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 255
The Genomics of Auditory Function and Disease

Shahar Taiber, Kathleen Gwilliam, Ronna Hertzano, and Karen B. Avraham p p p p p p p p p p 275
The Joubert–Meckel–Nephronophthisis Spectrum of Ciliopathies
Julie C. Van De Weghe, Arianna Gomez, and Dan Doherty p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 301
Mosaicism in Tumor Suppressor Gene Syndromes: Prevalence,
Diagnostic Strategies, and Transmission Risk
Jillian L. Chen, David T. Miller, Laura S. Schmidt, David Malkin,
Bruce R. Korf, Charis Eng, David J. Kwiatkowski, and Krinio Giannikou p p p p p p p p p p p 331
The Role of Telomeres in Human Disease
Mary Armanios p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 363
Decoding the Human Face: Progress and Challenges in

Understanding the Genetics of Craniofacial Morphology

Sahin Naqvi, Hanne Hoskens, Franziska Wilke, Seth M. Weinberg,
John R. Shaffer, Susan Walsh, Mark D. Shriver, Joanna Wysocka,
and Peter Claes p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 383
The Next Frontier in Noninvasive Prenatal Diagnostics: Cell-Free
Fetal DNA Analysis for Monogenic Disease Assessment
Lilian Pok Wa Zhong and Rossa W.K. Chiu p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 413
The Role of Genome Sequencing in Neonatal Intensive Care Units
Stephen F. Kingsmore and F. Sessions Cole p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 427
Advancing Pharmacogenomics from Single-Gene to
Preemptive Testing
Cyrine E. Haidar, Kristine R. Crews, James M. Hoffman, Mary V. Relling,
and Kelly E. Caudle p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 449
Diverse Molecular Mechanisms Underlying Pathogenic Protein
Mutations: Beyond the Loss-of-Function Paradigm
Lisa Backwell and Joseph A. Marsh p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 475
Five Priorities of African Genomics Research: The Next Frontier
Ambroise Wonkam, Nchangwi S. Munung, Collet Dandara, Kevin K. Esoh,
Neil A. Hanchard, and Guida Landoure p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 499
Mapping Human Reproduction with Single-Cell Genomics
Magda Marečková, Hassan Massalha, Valentina Lorenzi, and Roser Vento-Tormo p p p p p 523
Population Screening in Health Systems
Marc S. Williams p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 549
The UK Biobank: A Shining Example of Genome-Wide Association
Study Science with the Power to Detect the Murky Complications
of Real-World Epidemiology
Vanessa Y. Tan and Nicholas J. Timpson p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 569
Predicting Archaic Hominin Phenotypes from Genomic Data

Colin M. Brand, Laura L. Colbran, and John A. Capra p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 591
Equity in Genomic Medicine
Chanita Hughes Halbert p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 613
Ethical Guidance in Human Paleogenomics: New and Ongoing
Perspectives
Raquel E. Fleskes, Alyssa C. Bader, Krystal S. Tsosie, Jennifer K. Wagner,
Katrina G. Claw, and Nanibaa’ A. Garrison p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 627
Regulation of Molecular Diagnostics
Gail H. Javitt and Erik R. Vollebregt p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 653
Errata
An online log of corrections to Annual Review of Genomics and Human Genetics articles
may be found at http://www.annualreviews.org/errata/genom

Naqvi Et Al 2022 Decoding The Human Face Progress and Challenges in Understanding The Genetics of Craniofacial

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Naqvi Et Al 2022 Decoding The Human Face Progress and Challenges in Understanding The Genetics of Craniofacial

Uploaded by

Copyright:

Available Formats

Annual Review of Genomics and Human Genetics

Decoding the Human Face:

384 Naqvi et al.

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 385

ESTIMATES OF FACIAL FORM HERITABILITY

386 Naqvi et al.

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 387

RARE VARIATION IN CRANIOFACIAL MORPHOLOGY

388 Naqvi et al.

COMMON VARIATION IN CRANIOFACIAL SHAPE

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 389

HSPA9 GNPTAB FANCA KDM6A IRF6 TCF12 RIC1

Transporter, ion channel Mitochondria

Chromatin modification, remodeling

MASP1 SMARCA4 MSX2 VSX1 ROR2 PLCB4 EFTUD2

SEC31A CDCA7 RPS10

TBCE CDK13 RPS17

WDR37 PEX19 SMC3 TPM3 POLR1C GMNN

Table 1 GWASs of facial shape in healthy individuals

Genetics of Craniofacial Morphology and Disease

Genetics of Craniofacial Morphology and Disease

and craniosynostosis, are an interesting point of contrast to the architecture of normal-range

394 Naqvi et al.

terms of SNP effect size and significance (109).

Using Genome-Wide Association Studies to Understand Craniofacial Biology

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 395

INTERPLAY BETWEEN COMMON AND RARE VARIATION

396 Naqvi et al.

PAX3 TWIST1 ALX1 SOX9

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 397

cess directly inhibited by SMAD6.

A Multivariate Model Linking Craniofacial Shape Variation and Disease

398 Naqvi et al.

Cleft lip and palate Robin

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 399

Experimental Fine Mapping and Elucidating Target Genes

400 Naqvi et al.

variants in their endogenous context.

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 401

Epigenomic Maps from Relevant Cell Types

Mouse Models of Human Craniofacial Variation

402 Naqvi et al.

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 403

404 Naqvi et al.

syndromic 3D facial surface scans using 2D images. Sensors 20:3171

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 405

their roles in craniofacial development. Am. J. Med. Genet. A 155A:270–79

406 Naqvi et al.

disorders using deep learning. Nat. Med. 25:60–64

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 407

408 Naqvi et al.

Behav. Identity Sci. 4:163–72

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 409

410 Naqvi et al.

gene expression during cellular differentiation. Science 364:1287–90

defined in relation to underlying skeletal structures. Am. J. Orthod. 45:481–507

www.annualreviews.org • Genetics of Craniofacial Morphology and Disease 411

412 Naqvi et al.

Volume 23, 2022

A Journey Through Genetics to Biology

Vineet Bafna and Paul S. Mischel p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p29

Kenji Ito and Kenneth S. Zaret p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p53